Concise Encyclopedia of Robotics

concise encyclopedia of robotics - stan gibilisco

Get your own at Scribd or explore others:

Concise Encyclopedia of Robotics Other great robotics titles from TAB Electronics: Build Your Own Remote-Controlled Robot by David Shircliff Building Robot Drive Trains by Dennis Clarke and Michael Owings Combat Robots Complete by Chris Hannold Constructing Robot Bases by Gordon McComb Insectronics by Karl Williams Lego Mindstorms Interfacing by Don Wilcher Programming Robot Controllers by Myke Predko Robot Builder’s Bonanza by Gordon McComb Robot Builder’s Sourcebook by Gordon McComb Robots, Androids, and Animatrons by John Iovine Concise Encyclopedia of Robotics Stan Gibilisco McGraw-Hill New York Chicago San Francisco Lisbon London Madrid Mexico City Milan New Delhi San Juan Seoul Singapore Sydney Toronto of 1976, no part of this publication may be reproduced or distributed in any form or by any means, or stored in a database or retrieval system, without the prior written permission of the publisher. 0-07-141010-4 The material in this eBook also appears in the print version of this title: 0-07-142922-0. All trademarks are trademarks of their respective owners. Rather than put a trademark symbol after every occurrence of a trademarked name, we use names in an editorial fashion only, and to the benefit of the trademark owner, with no intention of infringement of the trademark. Where such designations appear in this book, they have been printed with initial caps. McGraw-Hill eBooks are available at special quantity discounts to use as premiums and sales promotions, or for use in corporate training programs. For more information, please contact George Hoare, Special Sales, at george_hoare@mcgraw-hill.com or (212) 9044069. TERMS OF USE its licensors reserve all rights in and to the work. Use of this work is subject to these terms. copy of the work, you may not decompile, disassemble, reverse engineer, reproduce, modify, create derivative works based upon, transmit, distribute, disseminate, sell, publish or sublicense the work or any part of it without McGraw-Hill’s prior consent. You may use the work for your own noncommercial and personal use; any other use of the work is strictly prohibited. Your right to use the work may be terminated if you fail to comply with these terms. THE WORK IS PROVIDED “AS IS”. McGRAW-HILL AND ITS LICENSORS MAKE NO GUARANTEES OR WARRANTIES AS TO THE ACCURACY, ADEQUACY OR COMPLETENESS OF OR RESULTS TO BE OBTAINED FROM USING THE WORK, INCLUDING ANY INFORMATION THAT CAN BE ACCESSED THROUGH THE WORK VIA HYPERLINK OR OTHERWISE, AND EXPRESSLY DISCLAIM ANY WARRANTY, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. McGraw-Hill and its licensors do not warrant or guarantee that the functions contained in the work will meet your requirements or that its operation will be uninterrupted or error free. Neither McGraw-Hill nor its licensors shall be liable to you or anyone else for any inaccuracy, error or omission, regardless of cause, in the work or for any damages resulting therefrom. McGraw-Hill has no responsibility for the content of any information accessed through the work. Under no circumstances shall McGraw-Hill and/or its licensors be liable for any indirect, incidental, special, punitive, consequential or similar damages that result from the use of or inability to use the work, even if any of them has been advised of the possibility of such damages. This limitation of liability shall apply to any claim or cause whatsoever whether such claim or cause arises in contract, tort or otherwise. DOI: 10.1036/0071410104 Want to learn more? We hope you enjoy this McGraw-Hill eBook! , If you d like more information about this book, its author, or related books and websites, please click here. To Samuel, Tim, and Tony from Uncle Stan This page intentionally left blank. For more information about this title, click here. Contents Foreword ix Introduction xi Acknowledgments xiii Concise Encyclopedia of Robotics and AI Suggested Additional References Index 1 351 353 vii This page intentionally left blank. Foreword W elcome to the high frontier of cognition and computation! You are about to dig into a uniquely interesting and important book. This is not a highly technical or abstruse guide to this often complex and difficult-to-understand subject. Rather, the book provides short, clear definitions and interpretations of the major concepts and rapidly emerging ideas in this dynamic field. The work includes numerous functional illustrations that help the general reader “see” the abstract robotics notions presented. I envision this book as an important introductory overview for the general interested reader and artificial intelligence (AI) hobbyist, and as a valuable backup and refresher for professional workers in the area. Because this book is all about terms, let me frame the effort by saying that it addresses two major, universal needs: cognition and computation. Cognition (kog-NISH-un) is literally “the act of knowing or awareness.” As you refer to definitions in this book, you should steadily gain knowledge and awareness of the robotics/AI topics presented. Most interestingly, it is your own central nervous system (brain and spinal cord), functionally connected to your eyes, that is allowing your natural intelligence (NI) to study this material and build a relevant cognition (mental awareness) of key robotics concepts. A thorough cognition and comprehension of robotics/AI terminology and concepts is becoming absolutely critical to all intelligent lay people worldwide. ix Foreword “Can human consciousness be duplicated electronically?” Stan Gibilisco thoughtfully asks us. “Will robots and smart machines ever present a danger to their makers? What can we reasonably expect from robotics and artificial intelligence in the next 10 years? In 50 years? In 100 years?” Such questions as these will be of ever-increasing importance for human cognition (NI) as machine computation or artificial intelligence continues to evolve rapidly. Let us coin a new word, computhink: a contraction of “computer-like modes or ways of human thinking.” A comprehensive introductory reference book such as this will help the NI of general readers to learn computhink. This will result in a better understanding and management of our powerful cousin, AI, for the greater benefit and education of all humankind. This book presents a thorough, basic, blissfully nonmathematical coverage of numerous electronic and mechanical concepts that is greatly needed worldwide. Stan has provided us with the essential vocabulary of machine computation for the twenty-first century, a vocabulary that many (not just a select few) people need to understand. THE HONORABLE DR. DALE PIERRE LAYMAN, PH.D. Founder and President, ROBOWATCH www.robowatch.org x Introduction T his is an alphabetical reference about robotics and artificial intelligence (AI) for hobbyists, students, and people who are just curious about these technologies. Computers and robots are here to stay. We depend on them every day. Often we don’t notice them until they break down. We will get more used to them, and more reliant on them, as the future unfolds. To find information on a subject, look for it as an article title. If your subject is not an article title, look for it in the index. This book is meant to be precise, but without too much math or jargon. It is written with one eye on today and the other eye on tomorrow. Illustrations are functional; they are drawn with the intention of showing, clearly and simply, how things work. Suggestions for future editions are welcome. STAN GIBILISCO Editor in Chief xi This page intentionally left blank. Acknowledgments I llustrations in this book were generated with CorelDRAW. Some clip art is courtesy of Corel Corporation, 1600 Carling Avenue, Ottawa, Ontario, Canada K1Z 8R7. xiii This page intentionally left blank. A ACOUSTIC PROXIMITY SENSOR An acoustic proximity sensor can be used by a robot to detect the presence of, and determine the distance to, an object or barrier at close range. It works based on acoustic wave interference. The principle is similar to that of sonar; but rather than measuring the time delay between the transmission of a pulse and its echo, the system analyzes the phase relationship between the transmitted wave and the reflected wave. When an acoustic signal having a single, well-defined, constant frequency (and therefore a single, well-defined, constant wavelength) reflects from a nearby object, the reflected wave combines with the incident wave to form alternating zones at which the acoustic energy adds and cancels in phase. If the robot and the object are both stationary, these zones remain fixed. Because of this, the zones are called standing waves. If the robot moves with respect to the object, the standing waves change position. Even a tiny shift in the relative position of the robot and the sensed object can produce a considerable change in the pattern of standing waves. This effect becomes more pronounced as the acoustic wave frequency increases, because the wavelength is inversely proportional to the frequency. The characteristics and effectiveness of an acoustic proximity sensor depend on how well the object or barrier reflects acoustic waves. A solid concrete wall is more easily detected than a sofa upholstered with cloth. The distance between the robot and the obstacle is a factor; in general, an acoustic proximity sensor works better as the distance decreases, and less well as the distance increases. The amount of acoustic noise in the robot’s work environment is also important. The higher the noise level, the more limited is the range over which the sensor functions, and the more likely are errors or false positives. Ultrasound waves provide exceptional accuracy at close range, in some cases less than 1 cm. Audible sound can allow the  Active Chord Mechanism (ACM) system to function at distances on the order of several meters. However, audible signals can annoy people who must work around the machine. Compare SONAR. See also PRESENCE SENSING and PROXIMITY SENSING. ACTIVE CHORD MECHANISM (ACM) An active chord mechanism (ACM) is a robot gripper that conforms to the shapes of irregular objects. An ACM is built something like the human backbone. A typical ACM consists of numerous small, rigid structures connected by hinges, as shown in the illustration. Hinges Rigid sections Active chord mechanism The precision with which an ACM can conform to an irregular object depends on the size and number of sections. The smaller the sections, the greater is the precision. An ACM exerts uniform pressure all along its length. This pressure can be increased or decreased, according to the required task. One application of ACMs is to position or arrange fragile objects without damaging them. Another application is the picking of fruits and vegetables. See also ROBOT GRIPPER. ACTIVE COOPERATION See COOPERATION.  AGV ACTUATOR An actuator is a device that moves one or more joints and operates the gripper or end effector in a robot arm. Simple actuators consist of electric motors and gears, cable drives, or chain drives. More sophisticated actuators involve the use of hydraulics, pneumatics, or magnetic interaction. Stepper motors are commonly used as robotic actuators. Some robot arms can function with a single actuator; others require two or more. The number of actuators necessary to perform a given task depends on the number of degrees of freedom, the number of degrees of rotation, and the coordinate geometry of the robot arm. See also CABLE DRIVE, CHAIN DRIVE, DEGREES OF FREEDOM, DEGREES OF ROTATION, END EFFECTOR, MOTOR, ROBOT ARM, ROBOT GRIPPER, and STEPPER MOTOR. ADAPTIVE SUSPENSION VEHICLE (ASV) An adaptive suspension vehicle (ASV) is a specialized robot that uses mechanical limbs to propel itself. It moves on several legs like a gigantic insect. This provides excellent stability and maneuverability. The ASV can carry several hundred kilograms, and moves at 2 to 4 m/s. The machine itself masses 2 to 3 metric tons. It is the size of a small truck, and it can carry a driver or rider. The design and construction of a robot with legs is considerably more difficult than that of a wheel-driven or track-driven robot, but there is a payoff: the ASV can move over much rougher terrain than any vehicle with wheels or a track drive. See also INSECT ROBOT and ROBOT LEG. ADHESION GRIPPER An adhesion gripper is a robot end effector that grasps objects by literally sticking to them. In its most primitive form, this type of gripper consists of a rod, sphere, or other solid object covered with two-sided tape. Velcro™ can also be used if the object(s) to be grasped are likewise equipped. A major asset of the adhesive gripper is the fact that it is simple. As long as the adhesive keeps its “stickiness,” it will continue to function without maintenance. However, there are certain limitations. The most significant is the fact that the adhesive cannot readily be disabled in order to release the grasp on an object. Some other means, such as devices that lock the gripped object into place, must be used. Compare ATTRACTION GRIPPER. AGV See AUTOMATED GUIDED VEHICLE.  Algorithm ALGORITHM An algorithm is a precise, step-by-step procedure by which a solution to a problem is found. Algorithms can usually be shown in flowchart form. All computer programs are algorithms. Robots perform specific tasks by following algorithms that tell them exactly where and when to move. In an efficient algorithm, every step is vital, even if it seems to sidetrack or backtrack. An algorithm must contain a finite number of steps. Each step must be expressible in digital terms, allowing a computer to execute it. Although the algorithm can contain loops that are iterated many times, the whole process must be executable in a finite length of time. Although no algorithm is infinitely complex, there are some that would require millions of years to be executed by a human being but can be done by computers in a few seconds. See also FLOWCHART. ALL-TRANSLATIONAL SYSTEM An all-translational system is a scheme in which the coordinate axes remain constant, or fixed, in an absolute sense as a robot moves. A common example is a system in three-dimensional (3-D) Cartesian coordinate geometry, in which the axes are defined as north/south, east/west, and up/down. An all-translational system in a given environment does not necessarily constitute an all-translational system in another environment. Consider a Cartesian system in which the x axis is north/south, the y axis is east/west, and the z axis is up/down. This is all-translational as defined in, and relative to, a small region on the Earth. However, this scheme loses its absoluteness with respect to the whole planet or the greater Universe, because the Earth is a rotating sphere, not a fixed Euclidean plane. In the absence of a set of physical objects for reference, an all-translational system can be maintained by inertial means. The gyroscope is the most common means of accomplishing this. See also CARTESIAN COORDINATE GEOMETRY and GYROSCOPE. ALTERNATIVE COMPUTER TECHNOLOGY Researchers in artificial intelligence (AI) have debated for years whether it is possible to build a machine with intelligence comparable to that of a human being. Some scientists think that alternative computer technology might provide a pathway in the quest for human-level AI. Digital processes Personal computers make use of digital computer technology. The operating language, known as machine language, consists of only two possible  Alternative Computer Technology states, the digits 1 and 0, represented by high and low electronic voltages. No matter how complex the function, graphic, or program, the workings of a digital computer can always be broken down into these two logic states. Digital computers can be made fast and powerful. They can work with huge amounts of data, processing it at many millions of digits per second. However, there are certain things that digital computers are not good at doing. Some researchers think that other approaches to computing deserve attention, even though digital technology has been successful so far. Analog processes Whereas a digital machine breaks everything down into discrete bits (binary digits), analog computer technology uses an entirely different approach. Think of the square root of 2. This cannot be represented as a ratio of whole numbers. A digital computer will calculate this and get a value of about 1.414. However, a decimal-number representation of the square root of 2 can never be exact. The best a digital machine can do is get close to its true value. The square root of 2 is the length of the diagonal of a square measuring 1 unit on a side. You can construct it with the tools of classical geometry (an analog art) and get an exact rendition. But you cannot use this in arithmetic as you use the numerical value 1.414. Thus, you sacrifice quantitative utility for qualitative perfection. Perhaps similar give-and-take will prove necessary in the quest to develop a computer that thinks like a human being. Analog concepts have been adapted to computer design; in fact, it was one of the earliest methods of computing. In recent years it has been largely ignored. Optics Visible light, infrared (IR), and ultraviolet (UV) offer interesting possibilities for the future of computer technology. In CD-ROM (compact disk, read-only memory), optical technology is used to increase the amount of data that can be stored in a given physical space. Tiny pits on a plastic diskette cause a laser beam to be reflected or absorbed at the surface. This allows encoding of many megabytes of data on a diskette less than 15 cm across. Data can be transmitted at extreme speeds, and in multiple channels, via lasers in glass fibers. This is known as fiber-optic data transmission, and is used in some telephone systems today. The wires in computers might someday be replaced by optical fibers. The digital logic states, now represented by electrical impulses or magnetic fields, would be represented  Alternative Computer Technology by light transmittivity instead. Certain materials change their optical properties very quickly, and can hold a given state for a long time. Atomic data As integrated circuit (IC) technology has advanced, more and more digital logic gates have been packed into less and less physical space. Also, with refinements in magnetic media, the capacity of hard disks and diskettes has been increasing. According to conventional science, the smallest possible data storage unit is a single atom or subatomic particle. Consider a magnetic diskette. Logic 1 might be represented by an atom “right side up,” with its magnetic north pole facing upward and its magnetic south pole facing downward. Then logic 0 would be represented by the same atom “upside down,” with the magnetic poles inverted. Another possibility is single-electron memory (SEM). An example of a SEM is a substance in which the presence of an excess electron in an atom represents logic 1, and the electrically neutral state of the atom represents logic 0. Some scientists think that computer chips might someday be grown in a laboratory, in a manner similar to the way experimental cultures of bacteria and viruses are grown. A name has even been coined for such a device: biochip. Nanotechnology As ICs get more circuitry packed into small packages, computer power increases. But it also becomes possible to make tinier and tinier computers. With molecular computer technology—the construction of ICs molecule by molecule rather than by etching material away from a chip—it might become possible to build computers so small that they can circulate inside the human body. Imagine antibody robots, controlled by a central computer, that are as small as bacteria. Suppose the central computer is programmed to destroy certain disease-causing organisms. Such a machine would be something like an artificial white blood cell. Nanotechnology is the field of research devoted to the development and programming of microscopic machines. The prefix nano- means one billionth (10 9 or 0.000000001). It also means “extremely small.” Computerized nanorobots might assemble larger computers, saving humans much of the work now associated with manufacturing the machines. Nanotechnology has already made it possible for you to wear a computer on your wrist, or even have a computer embedded somewhere in your body. See also BIOCHIP, INTEGRATED CIRCUIT, NANOCHIP, and NEURAL NETWORK.  Analogical Motion Neural networks Neural network technology uses a design philosophy that differs radically from that of conventional digital computers. Neural networks are good at spotting patterns, which is important for forecasting. Rather than working with discrete binary digits, neural networks work with the relationships among events. Unless there is a malfunction, a digital machine does precise things with data. This takes time, but the outcome is always the same if the input stays constant. This is not the case with a neural network. A neural network can work more quickly than a digital machine. To achieve speed, precision is sacrificed. Neural networks can learn from their mistakes. According to some scientists, this technology is a diversion and distraction from the proven mainstream; according to other scientists, it holds great promise. AMUSEMENT ROBOT An amusement robot is a hobby robot intended for entertainment or gaming. Companies sometimes use them to show off new products and to attract customers. They are common at trade fairs, especially in Japan. Although they are usually small in size, they often have sophisticated controllers. An example of an amusement robot is a mechanical mouse (not to be confused with the pointing device for a computer) that navigates a maze. The simplest such device bumps around randomly until it finds its way out by accident. A more sophisticated robot mouse moves along one wall of the maze until it emerges. This technique will work with most, but not all, mazes. The most advanced amusement robots include androids, or machines with a human appearance. Robots of this type can greet customers in stores, operate elevators, or demonstrate products at conventions. Some amusement robots can accommodate human riders. See also ANDROID and PERSONAL ROBOT. ANALOGICAL MOTION The term analogical motion refers to a variable or quantity that can have an infinite number of values within a certain range. This is in contrast to digital variables or quantities, which can have only a finite number of discrete values within a given range. Thus, analogical control is representative of so-called smooth or continuous motion. A person moving freely around a room, varying position to any point within a specific region, has the capability of analogical motion. The human arm can move to an infinite number of positions in a fluid and continuous way, within a certain region of space. This, too, is analog  Analytical Engine motion. Many robots, however, can move only to certain points along a line, on a plane, or in space. This motion is digital. Some robots can move in an analogical fashion, but the necessary hardware is generally more complicated than for digital motion. The illustration shows an example of analog motion in a plane. Compare DIGITAL MOTION. y x Analogical motion ANALYTICAL ENGINE The analytical engine was a primitive calculating machine designed by Charles Babbage in the nineteenth century. Babbage never completed the task of building this device to perfection, but the idea was to employ punched cards to make, and print out, calculations, in a manner similar to the earliest digital computers. Babbage is considered to be the first engineer to work on a true digital calculator. One of the main problems for Babbage was that electricity was not available. The machines had to use mechanical parts exclusively. These wore out with frequent, repetitive use. Another problem was that Babbage liked to dismantle things completely in order to start over with new designs,  Animism rather than saving his old machines to keep their shortcomings in mind when designing new ones. During the research-and-development phase of the analytical engine, some people thought that artificial intelligence (AI) had been discovered. The Countess of Lovelace even went so far as to write a program for the machine. Babbage’s machine represented a turning point in human attitudes toward machines. People began to believe that “smart machines” were not only possible in theory, but also practical. AND GATE See LOGIC GATE. ANDROID An android is a robot that has human form. A typical android has a rotatable head equipped with position sensors. Binocular machine vision allows the android to perceive depth, thereby locating objects anywhere within a large room. Speech recognition and speech synthesis can be included as well. Because of their quasi-human appearance, androids are especially suited for use where there are children. There are certain mechanical problems with design of humanoid robots. Biped robots are unstable. Even three-legged designs, while more stable, are two-legged whenever one of the legs is off the ground. Humans have an innate sense of balance, but this feature is difficult to program into a machine. Thus, an android usually propels itself by means of a wheel drive or track drive in its base. Elevators can be used to allow a rolling android to get from floor to floor in a building. The technology exists for fully functional arms, but the programming needed for their operation has not yet been made cost-effective for small robots. No android has yet been conceived, even on the trendiest drawing board, that can be mistaken for a person, as has been depicted in sciencefiction books and movies. Humanoid robots have enjoyed popularity, especially in Japan. One of the most famous was called Wasubot. It played an organ with the finesse of a professional musician. This robot became an idol at the Japanese show Expo ‘85. The demonstration showed that machines can be esthetically appealing as well as functional. See also PERSONAL ROBOT. ANIMISM People in some countries, notably Japan, believe that the force of life exists in things such as stones, lakes, and clouds, as well as in people, animals, and plants. This belief is called animism.  Anthropomorphism As early as the middle of the nineteenth century, a machine was conceived that was thought to be in some sense animate. This was Charles Babbage’s analytical engine. At that time, very few people seriously thought that a contraption made of wheels and gears could have life. However, today’s massive computers, and the promise of more sophisticated ones being built every year, have brought the question out of the realm of science fiction. Computers can do things that people cannot. For example, even a simple personal computer (PC) can figure out the value of (pi), the ratio of a circle’s circumference to its diameter, to millions of decimal places. Robots can be programmed to do things as complicated as figuring out how to get through a maze or rescue a person from a burning building. In recent years, programming has progressed to the point that computers can learn from their mistakes, so that they do not make any particular error more than once. This is one of the criteria for intelligence, but few Western engineers or scientists consider this, by itself, characteristic of life. ANTHROPOMORPHISM Sometimes, machines or other objects have characteristics that seem human-like to us. This is especially true of advanced computers and robots. We commit anthropomorphism when we think of a computer or robot as human. Androids, for example, are easy to anthropomorphize. Sciencefiction movies and novels often make use of anthropomorphisms. An example of anthropomorphism with respect to a computer occurs in the novel and movie 2001: A Space Odyssey. In this story, a spacecraft is controlled by “Hal,” a computer that becomes delusional and tries to kill the human astronauts. Some engineers believe that sophisticated robots and computers already have human qualities, because they can optimize problems and/or learn from their mistakes. Others, however, contend that the criteria for life are far more strict. Owners of personal robots sometimes think of the machines as companions. In that sense, such robots actually are like people, because it is possible to grow fond of them. See also PERSONAL ROBOT. ARM See ROBOT ARM. ARTICULATED GEOMETRY Robot arms can move in various different ways. Some can attain only certain discrete, or definite, positions, and cannot stop at any intermediate  Artificial Intelligence position. Others can move in smooth, sweeping motions, and are capable of reaching to any point within a certain region. One method of robot arm movement is called articulated geometry. The word “articulated” means “broken into sections by joints.” This type of robot arm resembles the arm of a human. The versatility is defined in terms of the number of degrees of freedom. There might, for example, be base rotation, elevation, and reach. There are several different articulated geometries for any given number of degrees of freedom. The illustration shows one scheme for a robot arm that uses articulated geometry. Shoulder Elbow Wrist Articulated geometry Other geometries that facilitate movement in two or three dimensions are defined under the titles CARTESIAN COORDINATE GEOMETRY, CYLINDRICAL COORDINATE GEOMETRY, DEGREES OF FREEDOM, POLAR COORDINATE GEOMETRY, and SPHERICAL COORDINATE GEOMETRY. ARTIFICIAL INTELLIGENCE The definition of what constitutes artificial intelligence (AI) varies among engineers. There is no universally accepted agreement on its exact meaning. The programming of robots can be divided into levels, starting with the least sophisticated and progressing to the theoretical, rather nebulous level of AI. The drawing shows a four-level programming scheme.  Artificial Stimulus Fourth level Third level Second level First level Artificial intelligence Tasks Complex motions Simple motions Artificial intelligence Artificial intelligence, at the top level, encompasses properties, behaviors, and tasks and involves robots with features such as the ability to: • • • • • • • • • • • • • Sense physical variables such as light and sound Generate high-resolution images (vision system) Develop a concept of reality (world model) Determine the optimum or most efficient course of action Learn from past mistakes Create a plan in a given situation, and then follow it through Modify a plan as changes occur in the environment Carry on two-way conversations with humans or other machines Infer solutions based on limited or incomplete information Develop new ways to solve old problems Search the knowledge base for specific facts or solutions Program themselves Improve their own designs Artificial intelligence is difficult to quantify; the most tempting standard is to compare “machine intelligence” with human intelligence. For example, a smart machine can be given an intelligence quotient (IQ) test similar to the tests designed to measure human intelligence. In this interpretation, the level of AI increases as a robot or computer becomes more “human-like” in its reactions to the world around it. Another scheme involves the use of games requiring look-ahead strategy, such as checkers or chess. ARTIFICIAL STIMULUS An artificial stimulus is a method of guiding a robot along a specified path. The automated guided vehicle (AGV), for example, makes use of a magnetic field to follow certain routes in its environment.  Assembly Robot Various landmarks can be used as artificial stimuli. It is not necessary to have wires or magnets embedded in the floor, as is the case with the AGV. A robot might be programmed to follow the wall on its right-hand (or left-hand) side until it reaches its destination, like finding its way out of a maze. The lamps in a hallway ceiling can be followed by light and direction sensors. The edge of a roadway can be followed by visually checking the difference in brightness between the road surface and the shoulder. Another way to provide guidance is to use a beacon. This can be an infrared (IR) or visible beam, or a set of ultrasound sources. With ultrasound, the robot can measure the difference in propagation time from different sources to find its position in an open space, if there are no obstructions. There are many ways that objects can be marked for identification. One method is bar coding, which is used for pricing and product identification in retail stores. Another is a passive transponder, of the type attached to merchandise to prevent shoplifting. See also AUTOMATED GUIDED VEHICLE, BAR CODING, BEACON, EDGE DETECTION, and PASSIVE TRANSPONDER. ASIMOV’S THREE LAWS In one of his early science-fiction stories, the prolific writer Isaac Asimov first mentioned the word “robotics,” along with three fundamental rules that all robots had to obey. The rules, now called Asimov’s three laws, are as follows. • A robot must not injure, or allow the injury of, any human being. • A robot must obey all orders from humans, except orders that would contradict the First Law. • A robot must protect itself, except when to do so would contradict the First Law or the Second Law. Although these rules were first coined in the 1940s, they are still considered good standards for robotic behavior. ASSEMBLY ROBOT An assembly robot is any robot that assembles products, such as cars, home appliances, or electronic equipment. Some assembly robots work alone; most are used in automated integrated manufacturing systems (AIMS), doing repetitive work at high speed and for long periods of time. Many assembly robots take the form of robot arms. The type of joint arrangement depends on the task that the robot must perform. Joint arrangements are named according to the type of coordinate system they follow. The complexity of motion in an assembly robot is expressed in terms of the number of degrees of freedom.  Attraction Gripper To do its work properly, an assembly robot must have all the parts it works with placed in exactly the correct locations. This ensures that the robot can pick up each part in the assembly process, in turn, by moving to the correct set of coordinates. There is little tolerance for error. In some assembly systems, the various components are labeled with identifying tags such as bar codes, so the robot can find each part by zeroing in on the tag. See also CARTESIAN COORDINATE GEOMETRY, CYLINDRICAL COORDINATE GEOMETRY, DEGREES OF FREEDOM, POLAR COORDINATE GEOMETRY, ROBOT ARM, and SPHERICAL COORDINATE GEOMETRY. ATTRACTION GRIPPER An attraction gripper is a robot end effector that grasps objects by means of electrical or magnetic attraction. Generally, magnets are used; either permanent magnets or electromagnets will serve the purpose. Electromagnets offer the advantage of being on/off controllable, so an object can be conveniently released without its having to be secured by some external means. Permanent magnets, conversely, offer the advantage of a minimal maintenance requirement. Like the adhesive gripper, the attraction gripper is fundamentally simple. There are two primary problems with this type of end effector. First, in order for a magnetic attraction gripper to work, the object it grasps must contain a ferromagnetic material such as iron or steel. Second, the magnetic field produced by the end effector can permanently magnetize the objects it handles. In some cases this is not a concern, but in other instances it can cause trouble. Compare ADHESION GRIPPER. ATTRACTIVE RADIAL FIELD See POTENTIAL FIELD. AUTOMATED GUIDED VEHICLE An automated guided vehicle (AGV) is a robot cart that runs without a driver. The cart has an electric engine and is guided by a magnetic field, produced by a wire on or just beneath the floor (see the illustration). Alternatively, an AGV can run on a set of rails. In automated systems, AGVs are used to bring components to assembly lines. AGVs can also serve as attendants in hospitals, bringing food and nonessential items to patients, or as mechanical gophers to perform routine chores around the home or office. There has been some talk about making automobiles into AGVs that follow wires embedded in the road pavement. This would take part of the driver’s job away, letting computers steer the vehicle and adjust its speed.  Automaton Electromagnet Floor Current-carrying wire Automated guided vehicle Each car would have its own individual computer. In a city, traffic would be overseen by one or more central computers. In the event of computer failure, all traffic would stop. This would practically eliminate accidents. Whether the public would accept this sort of system on a general scale remains to be seen. AUTOMATED HOME See SMART HOME. AUTOMATION The term automation refers to a system in which some or all of the processes are done by machines, particularly robots. Assets of automation include the following: • • • • Robots work fast. Robots are precise. Robots are reliable if they are well designed and maintained. Robots are capable of enormous physical strength. Advantages of human operators over robots include these facts: • • • • People can solve some problems that machines cannot. People have greater tolerance for confusion and error. Humans can perform certain tasks that robots cannot. Humans are needed to supervise robotic systems. AUTOMATON An automaton is a simple robot that performs a task or set of tasks without sophisticated computer control. Automata have been around for over 200 years.  Autonomous Robot An early example of an automaton was the “mechanical duck” designed by J. de Vaucanson in the eighteenth century. It was used to entertain audiences in Europe. It made quacking sounds and seemed to eat and drink. Vaucanson used the robot act to raise money for his work. Every December, certain ambitious people build holiday displays in their yards, consisting of machines in the form of people and animals. These machines have no “brains,” because they simply follow mechanical routines. Although they are fun to observe, these devices lack precision, and the motions they can make are limited. Some of these machines may look like androids, but are actually no more than moving statues. Compare ANDROID. AUTONOMOUS ROBOT An autonomous robot is self-contained, housing its own controller, and not depending on a central computer for its commands. It navigates its work environment under its own power, usually by rolling on wheels or a track drive. Robot autonomy might at first seem like a great asset: if a robot functions by itself in a system, then when other parts of the system fail, the robot will keep working. However, in systems where many identical robots are used, autonomy is inefficient. It is better from an economic standpoint to put programs in one central computer that controls all the robots. Insect robots work this way. Simple robots, like those in assembly lines, are not autonomous. The more complex the task, and the more different things a robot must do, the more autonomy it can have. The most advanced autonomous robots have artificial intelligence (AI). See also ANDROID and INSECT ROBOT. AXIS INTERCHANGE Axis interchange is the transposition of coordinate axis in a robotic system that uses Cartesian coordinate geometry. Axis interchange can involve two axes, or all three. The illustration shows an example in which the left/right (normally x) and up/down (normally z) axes are transposed. This is not the only way in which the left/right versus up/down interchange can take place; one or both axes might also be inverted. Clearly, there are numerous possibilities for axis interchange in a three-dimensional Cartesian system. Axis interchange can produce useful variations in robot movements. A single-motion programming scheme can result in vastly different work envelopes and motion patterns, depending on how the axes are defined. No matter how the axes are transposed, however, there is always a one-to-one correspondence between the points in both work envelopes, provided the motion programming is done properly.  Axis Inversion +z -y -x +x Normal +y -z +x -y -z +z +y -x Axis interchange Left/right and up/down axes interchanged Depending on the type of robotic system used, axis interchange can alter or limit the work envelope. Certain position points, or certain types of motion, that are possible in one coordinate scheme might be impossible in the other. See also AXIS INVERSION, CARTESIAN COORDINATE GEOMETRY, and WORK ENVELOPE. AXIS INVERSION Axis inversion is a reversal in the orientation of one or more coordinate axes in a robotic system that uses Cartesian coordinate geometry.  Axis Inversion When robotic motions are programmed using the Cartesian (or rectangular) scheme, the difference between right-handed and left-handed operations consists only of the reversal, or inversion, of the coordinates in one of the axes. Generally, the left/right axis in a Cartesian scheme is the x axis. The reversal of the coordinates in this axis is a form of singleaxis inversion. The illustration shows two three-dimensional Cartesian coordinate grids. In the top example, a right-handed scheme is depicted. The lower drawing shows the left-handed equivalent. The coordinate designations +z -y -x +x Right-handed +y -z +z -y +x -x Left-handed +y -z Axis inversion  Azimuth-Range Navigation are identical, except that they are mirror images with respect to the x axis. All the divisions represent the same unit distance in either case. While left and right are reversed in this example, the senses of up/down and forward/backward remain the same. In some systems, it is necessary to invert two, or even all three, axes to obtain the desired robot motion. These schemes can be called dual-axis inversion or triple-axis inversion. See also AXIS INTERCHANGE and CARTESIAN COORDINATE GEOMETRY. AZIMUTH-RANGE NAVIGATION Electromagnetic (EM) or acoustic waves reflect from various objects. By ascertaining the directions from which transmitted EM or acoustic signals are returned, and by measuring the time it takes for pulses to travel from the transmitter location to a target and back, it is possible for a robot to locate objects within its work environment. The ongoing changes in the azimuth (compass bearing) and range (distance) information for each object in the work environment can be used by the robot controller for navigation. 0 Azimuth 270 Robot Range 90 Target 180 Azimuth-range navigation  Azimuth-Range Navigation A classical azimuth-range navigation system is conventional radar, which consists of a transmitter, a highly directional antenna, a receiver, and a display. The transmitter produces EM microwave pulses that are propagated in a narrow beam. The EM waves strike objects at various distances. The greater the distance to the target, the longer is the delay before the echo is received. The transmitting antenna is rotated so that all azimuth bearings can be observed. The basic configuration of an azimuth-range scheme is shown in the illustration. The robot is at the center of the display. Azimuth bearings are indicated in degrees clockwise from true north, and are marked around the perimeter. The distance, or range, is indicated by the radial displacement. Some azimuth-range systems can detect changes in the frequencies of returned EM or acoustic pulses resulting from Doppler effect. These data are employed to measure the speeds of approaching or receding objects. The robot controller can use this information, along with the position data afforded by the azimuth-range scheme, to navigate in complex environments. See also RADAR.  B BACK LIGHTING In a robotic vision system, back lighting refers to illumination of objects in the work environment using a light source generally in line with, but more distant than, the objects. The light from the source therefore does not reflect from the surfaces of the objects under observation. Back lighting is used in situations where the surface details of observed objects are not of interest or significance to the robot, but the shape of the projected image is of importance. Back lighting is also advantageous in certain situations involving translucent or semitransparent objects whose internal structure must be analyzed. Light rays passing through a translucent or semitransparent object can reveal details that front lighting or side lighting cannot. Compare FRONT LIGHTING and SIDE LIGHTING. BACK PRESSURE SENSOR A back pressure sensor is a device that detects, and measures, the amount of torque that a robot motor applies at any given time. The sensor produces a signal, usually a variable voltage called the back voltage, that increases as the torque increases. The back voltage is used as negative feedback to limit the torque applied by the motor. When a robot motor operates, it encounters mechanical resistance called back pressure. This resistance depends on various factors, such as the weight of an object being lifted, or the friction of an object as it is moved along a surface. The torque is a direct function of mechanical resistance. As the torque increases, so does the back pressure the motor encounters. Conversely, as the back pressure increases, so does the motor torque necessary to produce a given result. Back pressure sensors and feedback systems are used to limit the amount of force applied by a robot gripper, arm, drill, hammer, or other device. This can prevent damage to objects being handled by the robot. It also helps to ensure the safety of people working around the robot. The accompanying illustration is a functional block diagram of the operation  Backward Chaining Torque Robot arm Force Motor Back pressure Load Sensor Signal Back pressure sensor Resistance of a back pressure sensor and the associated negative feedback loop that governs the applied torque. See also ROBOT ARM and ROBOT GRIPPER. BACKWARD CHAINING Backward chaining is a logical process that can be used in artificial intelligence (AI). Rather than working with data that have been supplied in advance, the computer requests data as it goes along. In this way, the computer gets only the information it needs to solve a problem. No memory is wasted in storing unnecessary data. Backward chaining is especially useful in expert systems, which are programs designed to help solve specialized problems in unfamiliar fields. A good example is a medical-diagnosis program. Backward chaining can also be of use in electronic troubleshooting, weather forecasting, cost analysis, and even police detective work. Compare FORWARD CHAINING. See also EXPERT SYSTEM. BALLISTIC CONTROL Ballistic control is a form of robotic motion control in which the path, or trajectory, of the device is calculated or programmed entirely in advance. Once the path has been determined, no further corrections are made. The term derives from the similarity to ballistics calculations for aiming guns and missiles.  Bandwidth The main assets of ballistic control are simplicity and moderate cost. A robotic manipulator with ballistic control does not have to carry sensors; a mobile robot with ballistic control does not need its own on-board computer. The main limitation is the fact that ballistic control does not allow for rapid, localized, or unexpected changes in the work environment. Compare CLOSED-LOOP CONTROL. BANDWIDTH Bandwidth refers to the amount of frequency space, or spectrum space, that a signal requires in order to be transmitted and received clearly. Bandwidth is generally defined as the difference in frequency between the two half-power points in a transmitted or received data signal, as shown in the illustration. All signals have a finite, nonzero bandwidth. No signal can be transmitted in an infinitely tiny slot of spectrum space. In general, the bandwidth of a signal is proportional to the speed at which the data are sent and received. In digital systems, data speed is denoted in bits per second (bps), kilobits per second (kbps), megabits per second (Mbps), or gigabits per second (Gbps), where 1 kbps = 1000 bps = 103 bps 1 Mbps = 1000 kbps = 106 bps 1 Gbps = 1000 Mbps = 109 bps Relative signal power Half-power points Relative Frequency Bandwidth Bandwidth  Bar Coding As the allowable bandwidth is increased, the maximum data speed increases in direct proportion. As the allowable bandwidth is restricted, the maximum data speed decreases in direct proportion. BAR CODING Bar coding is a method of labeling objects. Bar-code labels or tags are used extensively in retail stores for pricing and identifying merchandise. A bar-code tag has a characteristic appearance, with parallel lines of varying width and spacing (see illustration). A laser-equipped device scans the tag, retrieving the identifying data. The reading device does not have to be brought right up to the tag; it can work from some distance away. Bar coding Bar-code tags are one method by which objects can be labeled so that a robot can identify them. This greatly simplifies the recognition process. For example, every item in a tool set can be tagged using barcode stickers, with a unique code for each tool. When a robot’s controller tells the machine that it needs a certain tool, the robot can seek out the appropriate tag and carry out the movements according to the program subroutine for that tool. Even if the tool gets misplaced, as long as it is within the robot’s work envelope or range of motion, it can easily be found. See also PASSIVE TRANSPONDER. BATTERY POWER See ELECTROCHEMICAL POWER and SOLAR POWER. BEACON A beacon is a device used to help robots navigate. Beacons can be categorized as either passive or active. A mirror is a good example of a passive beacon. It does not produce a signal of its own; it merely reflects light beams that strike it. The robot requires a transmitter, such as a flashing lamp or laser beam, and a receiver,  Biased Search such as a photocell. The distance to each mirror can be determined by the time required for the flash to travel to the mirror and return to the robot. Because this delay is an extremely short interval of time, high-speed measuring apparatus is needed. An example of an active beacon is a radio transmitter. Several transmitters can be put in various places, and their signals synchronized so that they are all exactly in phase. As the robot moves around, the relative phase of the signals varies. Using an internal computer, the robot can determine its position by comparing the phases of the signals from the beacons. With active beacons, the robot does not need a transmitter, but the beacons must have a source of power and be properly aligned. See also ARTIFICIAL STIMULUS. BEHAVIOR In robotics, behavior refers to the processing of sensor data into specific motions, sequences of motions, or tasks. There are three main types: reflexive behavior, reactive behavior, and conscious behavior. Reflexive behavior is the simplest and fastest form of robotic behavior. Sensors can be, and often are, connected directly to manipulators, propulsion systems, or other mechanical devices. An electric eye that triggers an intrusion alarm is a good example of a device that employs reflexive behavior. When the light beam is broken, an electric current is interrupted, and this actuates an electronic switch that applies power to an acoustic emitter. Reactive behavior involves a primitive sort of machine intelligence; the extent or nature of the action varies over a range that depends on one or more parameters in the work environment. An example of reactive behavior is the operation of a back pressure sensor, in which the amount of torque applied by a robotic arm or end effector varies depending on the mechanical resistance offered by the manipulated object. Conscious behavior involves artificial intelligence (AI), in which a robot controller performs complex tasks such as playing chess or making choices that depend on multiple factors that cannot be predicted. BIASED SEARCH A biased search is an analog method by which a mobile robot can find a destination or target, by first looking off to one side and then “zeroing in.” The illustration shows a biased-search scheme that a boater might use on a foggy day. At some distance from the shoreline, the boater cannot see the dock, but has a reasonably good idea of where it is. Therefore, an approach is deliberately made well off to one side (in this case, to the left) of the dock. When the shore comes into view, the boater turns to the right and follows it until the dock is found.  Binary Number System Shoreline Dock Shoreline sighted Boat follows shoreline Initial course Initial position Biased search For a robot to use this technique effectively, it must have some familiarity with its environment, just as the boater knows roughly where the dock will be. This is accomplished by means of task-level programming, a primitive form of artificial intelligence (AI). Compare BINARY SEARCH. See also TASK-LEVEL PROGRAMMING. BINARY NUMBER SYSTEM See NUMERATION. BINARY SEARCH In a digital computer, a binary search, also called a dichotomizing search, is a method of locating an item in a large set of items. Each item in the set is given a number key. The number of keys is always a power of 2. Therefore, when it is repeatedly divided into halves, the end result is always a single key. If there are 16 items in a list, for example, they might be numbered 1 through 16. If there are 21 items, they can be numbered 1 through 21, with the numbers 22 through 32 as “dummy” keys (unoccupied). The desired number key is first compared with the highest number in the list. If the desired key is less than half the highest number in the list, then the first half of the list is accepted, and the second half is rejected. If the  Binaural Machine Hearing desired key is larger than half the highest number in the list, then the second half of the list is accepted, and the first half is rejected. The process is repeated, each time selecting half of the list and rejecting the other half, until only one item remains. This item is the desired key. The illustration shows an example of a binary search to choose one item from a list of 21. Keys are indicated by filled-in squares, except for the desired key, 21, which is indicated by a shaded circle. “Dummy” keys are shown as open squares. Compare BIASED SEARCH. 1. 2. 3. 4. 5. 6. Binary search BINAURAL MACHINE HEARING Binaural machine hearing utilizes two sound transducers, spaced a certain minimum distance from each other, to determine the direction from which acoustic waves are coming. This is done by comparing the relative phase and/or the relative loudness of the incoming wavefronts at the transducers. The human ear/brain system processes acoustic information to a high degree of exactitude, allowing a person to locate a sound source with remarkable accuracy even when the source cannot be seen. When equipped with sensitive transducers, a circuit called a phase comparator, and a sophisticated controller, a robot can do the same. In binaural machine hearing, two sound transducers are positioned on either side of a robot’s “head.” The phase comparator measures the relative phase and intensity of the signals from the two transducers. These data are sent to the controller, letting the robot determine, with  Binocular Machine Vision Incoming sound waves Left transducer Robot's “head” Right transducer Amplifier Phase comparator Amplifier Robot controller Binaural machine hearing certain limitations, the direction from which the sound is coming (see the illustration). If the system is confused, the robot head can turn until the confusion is eliminated and a meaningful bearing is obtained. BINOCULAR MACHINE VISION Binocular machine vision is the analog of binocular human vision. It is sometimes called stereoscopic vision. In humans, binocular vision allows perception of depth. With one eye, that is, with monocular vision, a human can infer depth to some extent on  Bin Picking Problem the basis of perspective. Almost everyone, however, has had the experience of being fooled when looking at a scene with one eye covered or blocked. A nearby pole and a distant tower might seem to be near each other, when in fact they are hundreds of meters apart. In a robot, binocular vision requires a sophisticated microprocessor. The inferences that humans make, based on what the two eyes see, are exceedingly complicated. The illustration shows the basic concept for binocular machine vision. Left eye sees: Right eye sees: Controller Machine sees Depth and perspective Binocular machine vision Of primary importance for good binocular robot vision are the following: • High-resolution visual sensors • A sophisticated robot controller • Programming in which the robot acts on commands, based on what it sees See also VISION SYSTEM. BIN PICKING PROBLEM A bin picking problem is a challenge presented to a robotic vision system in which the machine must choose a specific object from a group of objects. Basic machine vision systems can see only the outlines of objects; depth perception is lacking. As viewed from different angles by such a vision system, the appearance of an object can vary dramatically.  Biochip Side view Top view View from an angle Bin picking problem The illustration denotes the example of a cylindrical drinking glass. When seen from exactly side-on, it looks like a rectangle and its interior (left). From the top or the bottom, it looks like a circle and its interior (center). From an intermediate angle, it has a shape similar to that shown at right. The problem of object recognition is compounded when a certain object must be picked from a bin containing many other objects. Some, most, or all of the desired object can be obscured by other objects. One of the greatest challenges in developing artificial intelligence (AI) is giving robots the ability to solve these kinds of problems. One way to help a robot select items from a bin is to give each item a code. This can be done by means of bar coding or passive transponders. See also BAR CODING and PASSIVE TRANSPONDER. BIOCHIP A biochip is an integrated circuit (IC) fabricated with, or from, living matter by means of biological processes. The term has also been suggested for ICs manufactured using techniques that mimic the way nature puts atoms together. It has been suggested that the human brain is actually nothing more than a sophisticated computer. Any digital computer, no matter how complex, is always built up from individual logic gates. Whether the same can be said for the human brain remains to be seen. Nature assembles a brain (or any other living matter) by putting protons, neutrons, and electrons together in specific, predetermined patterns. Every proton is identical to every other proton; the same is true for neutrons and electrons. The building blocks are simple. It is the way they are combined that is complicated. Based on these premises, it is reasonable to suppose that a computer can be “grown” that is as smart as a human brain. Some researchers look at  Biomechatronics the way nature builds things to get ideas for the construction of improved ICs. The ultimate goal is a biochip that sprouts and evolves as a plant, from a specially engineered “seed.” See also INTEGRATED CIRCUIT. BIOLOGICAL ROBOT A biological robot is a hypothetical machine derived by cloning from living organisms, and grown in a laboratory environment to perform a specific function or set of functions. Research has been done in this field, although true biological robots have not yet been fabricated or grown. Biological robots have served as characters in science-fiction stories. The possibilities posed by this notion are limited only by the imagination. There are ethical questions and problems in biological-robot research. These issues are of such grave concern that some scientists refuse to work in this field. See also CYBORG. BIOMECHANISM A biomechanism is a mechanical device that simulates the workings of some part of a living body. Examples of biomechanisms are mechanical hands, arms, and legs, known in the medical field as prostheses. Especially, the term applies to robotic devices that not only perform the functions of their living counterparts, but look like them. The term biomechanism can also be used in reference to some body function. Thus, one might talk about the structure of a forearm and hand, calling it a biomechanism. The human anatomy has, in fact, proven to be an excellent model for the design of robotic devices. See also BIOMECHATRONICS. BIOMECHATRONICS The word biomechatronics is a contraction of the words biology, mechanics, and electronics. The field of biomechatronics is part of the larger realms of robotics and artificial intelligence (AI). Specifically, biomechatronics involves electronic and mechanical devices that duplicate human body parts and their functions. Biomechatronics has received more attention in Japan than in the United States. In Japan, some robot researchers attack their problems with religious zeal. Not only would Japanese robotics engineers like to build robots that can do all the things people can do, but some want their robots to look like people, too. The ultimate biomechatronic device is an android. Scientists generally agree that an intelligent android will not be developed for many years.  Biped Robot The problem of making androids can be approached from two directions. On the one hand, biological robots might be grown in laboratories by a process of cloning. This idea is clouded by profound ethical issues. On the other hand, engineers can try to build a mechanical robot with the dexterity and intelligence of a human being. This notion, too, brings up ethical questions, but to a lesser degree. See also ANDROID, BIOCHIP, BIOLOGICAL ROBOT, BIOMECHANISM, and CYBORG. BIPED ROBOT A biped robot is a robot with two legs that are used for support and propulsion. Usually, but not always, such robots have arms and a head, so they are androids. Physically, biped robots are unstable unless equipped with specialized balancing systems. Humans can manage with two legs because the brain and inner ear together constitute a feedback system that provides a good sense of balance. The human sense of balance can be duplicated electromechanically, but the designs are sophisticated and expensive. Robots that use legs for propulsion generally have four or six legs, because these designs offer better inherent stability than the biped scheme. See also INSECT ROBOT, QUADRUPED ROBOT, and ROBOT LEG. BIT-MAPPED GRAPHICS In a robotic vision system, an image can be assembled from thousands of tiny square elements. The smaller the elements, called pixels, the more detail the image can show for a given image size. Images made this way are bit-mapped graphics, also known as raster graphics. On a computer display, the image you see is a pattern of pixels in a fine, interwoven mesh. You can observe these pixels if you dim your monitor so you can hardly see the image (this is important!) and then look closely at it through a high-powered magnifying lens. A computer stores bitmapped graphic images as a vast array of logic highs and lows (ones and zeros). To obtain an image from this array of bits, the computer employs a function called a bit map. Bit-mapped graphics always produce approximations of scenes or objects. This is because each pixel is a square, and can take only certain digital values. If the number of pixels in an image is extremely large, the approximation is a good representation of reality in most instances. However, the detail obtainable with bit-mapped graphics is always limited by the image resolution. Bit-mapped graphics produce an artifact called jaggies or aliasing, a peculiar “digitized” appearance in the edges of rendered objects. Vertical and horizontal lines look all right, but curves and diagonals are roughened with “saw teeth.” To some extent this can be reduced by means of antialiasing  Blackboard System software or photocopy reduction, but a better way is to use object-oriented graphics. Compare OBJECT-ORIENTED GRAPHICS. See also COMPUTER MAP and VISION SYSTEM. BLACKBOARD SYSTEM A blackboard system incorporates artificial intelligence (AI) to help a computer recognize sounds or images. The incoming signal is digitized using an analog-to-digital converter (ADC). The digital data is input to a read/write memory circuit called the blackboard. Then the digital data is evaluated by various specialty programs. The overall scheme is depicted in the diagram. Robot controller S S Read/write memory S S = Specialty programs S ADC Analog data from sensors Blackboard system  Bladder Gripper For speech recognition, specialties include vowel sounds, consonant sounds, grammar, syntax, context, and other variables. For example, a context specialty program might determine whether a speaker means to say “weigh” or “way,” or “two,” “too,” or “to.” Another program lets the controller know when a sentence is finished and the next sentence is to begin. Another program can tell the difference between a statement and a question. Using the blackboard as their forum, the specialty circuits “debate” the most likely and logical interpretations of what is heard or seen. A “referee” called a focus specialist mediates. For object recognition, specialties might be shape, color, size, texture, height, width, depth, and other visual cues. How does a computer know if an object is a cup on a table, or a water tower a mile away? Is that a bright lamp, or is it the sun? Is that biped thing a robot, a mannequin, or a person? As with speech recognition, the blackboard serves as a debating ground. See also OBJECT RECOGNITION and SPEECH RECOGNITION. BLADDER GRIPPER A bladder gripper or bladder hand is a specialized robotic end effector that can be used to grasp, pick up, and move rod-shaped or cylindrical objects. The main element of the gripper is an inflatable, donut-shaped or cylindrical sleeve that resembles the cuff commonly used in blood pressure measuring apparatus. The sleeve is positioned so it surrounds the object to be gripped, and then the sleeve is inflated until it is tight enough to accomplish the desired task. The pressure exerted by the sleeve can be measured and regulated using force sensors. Bladder grippers are useful in handling fragile objects. However, they do not operate fast, and they can function only with objects within a rather narrow range of physical sizes. See also ROBOT GRIPPER. BONGARD PROBLEM The Bongard problem, named after its inventor, is a method of evaluating how well a robotic vision system can differentiate among patterns. Solving such problems requires a certain level of artificial intelligence (AI). An example of a Bongard problem is shown in the illustration. There are two groups of six boxes. The contents of the boxes on the left all have something in common; those on the right have the same characteristic in common, but to a different degree, or in a different way. To solve the problem, the vision system (or you) must answer three questions: • What do the contents of the boxes to the left of the heavy, vertical line have in common? • What do the contents of the boxes to the right of the line have in common?  Boolean Algebra X + + * Bongard problem • What is the difference between the contents of the boxes on opposite sides of the heavy, vertical line? In this case, the boxes on the left contain four dots or straight lines each; those on the right contain five dots or straight lines each. The difference between the boxes on the left and those on the right, therefore, is in the number of dots or straight lines they contain. See also OBJECT RECOGNITION. BOOLEAN ALGEBRA Boolean algebra is a system of mathematical logic using the numbers 0 and 1 with the operations AND (multiplication), OR (addition), and NOT (negation). Combinations of these operations are NAND (NOT AND) and NOR (NOT OR). Boolean functions are used in the design of digital logic circuits. In Boolean algebra, X AND Y is written XY or X*Y. NOT X is written with a line or tilde over the quantity, or as a minus sign followed by the quantity. X OR Y is written X+Y. The first table shows the values of these functions, where 0 indicates “falsity” and 1 indicates “truth.” The statements Boolean algebra: basic operations X Y X X*Y 0 0 1 0 0 1 1 1 0 1 1 0 0 0 0 1 X+Y 0 1 1 1  Branching Boolean algebra: theorems Equation X+0=X X*1=X X+1=1 X*0=0 X+X=X X*X=X ( X) = X X + ( X) = X X * ( X) = 0 X+Y=Y+X X*Y=Y*X X + (X * Y) = X X * ( Y) + Y = X + Y X + Y + Z = (X + Y) + Z = X + (Y + Z) X * Y * Z = (X * Y) * Z = X * (Y * Z) X * (Y + Z) = (X * Y) + (X * Z) (X + Y) = ( X) * ( Y) (X * Y) = ( X) + ( Y) Double negation Contradiction Commutativity of OR Commutativity of AND Name (if applicable) OR identity AND identity Associativity of OR Associativity of AND Distributivity DeMorgan’s theorem DeMorgan’s theorem on either side of the equal sign are logically equivalent. The second table shows several logic equations. These are facts, or theorems. Boolean theorems can be used to analyze complicated logic functions. See also LOGIC GATE. BRANCHING Branching refers to routines, or programs, that have points at which an intelligent robot controller must select among alternatives. Consider a robot on an assembly line that makes cars. The robot’s job is to insert hubcaps in the two right-side wheels. (An identical robot does the same job on the left side.) Suppose that 20 percent of the cars are fitted with gold-colored (G) hubcaps; the rest are fitted with silver-colored (S) ones. The robot should insert hubcaps in the following sequence: SS SS SS SS GG SS SS SS SS GG SS SS…, and so on. Every fifth pair of hubcaps is gold. Each time a hubcap pair is to be inserted, the computer must make a choice. Thus, the routine is at a branch point for every hubcap pair. Every  Branching fifth time the choice must be made, the robot controller chooses gold hubcaps. Otherwise, it chooses silver ones. This sequence is programmed into the controller. The logical process proceeds something like the flowchart in the accompanying figure. Identify car number (1,2,3,...) No Divisible by 5? Yes Install SS Install GG Add 1 to car number Branching Suppose a glitch occurs, in which the robot controller or hardware omits or overlooks a single hubcap. This will throw off the robot’s perception of the sequence of cars, so it thinks a new car has arrived with each set of rear wheels. Shortly, the front wheel of a car will get a silver hubcap and the rear wheel of the same car will get a gold one. The next car will get a gold hubcap on the front wheel and a silver one on the rear wheel. The repercussions will be repeated down the line over and over,  Bumper messing up two out of every five cars, or 40 percent of the automobiles coming off the assembly line. See also EXPERT SYSTEM. BUMPER See PROXIMITY SENSING. BURN-IN Before any electronic or electromechanical system is put to use, it should undergo a burn-in process. This usually involves running the system continuously for hours, days, or weeks. In some cases, a faulty system fails shortly after it is put online. In many instances, however, failure does not occur until a considerable time has passed. Intermittent failures might not manifest themselves until many hours have passed with continuous supervision. The burn-in process can weed out systems with early-failure problems, minimizing real-time failures. See also QUALITY ASSURANCE AND CONTROL.  C CABLE DRIVE A cable drive is a method of transferring mechanical energy in a robotic system from an actuator to a manipulator or end effector. This type of drive can also be used in wheel-drive propulsion systems and in certain indicating devices. The system consists of a cable, or cord, and a set of pulleys. The main asset of the cable drive is its simplicity. The principal limitation is the fact that the cable can slip on the wheels or pulleys, and over time, the cable can degenerate, and ultimately break without warning. Anyone who has been stranded on a highway because of a failed automotive fan belt can attest to the problems this can cause. Compare CHAIN DRIVE. CAPACITIVE PRESSURE SENSOR A capacitive pressure sensor consists of two metal plates separated by a layer of nonconductive (dielectric) foam. The resulting variable capacitor is connected in parallel with an inductor; the inductance/capacitance (LC) circuit determines the frequency of an oscillator. If an object strikes the sensor, the plate spacing momentarily decreases. This increases the capacitance, causing a drop in the oscillator frequency. When the object moves away from the transducer, the foam springs back, the plates return to their original spacing, and the oscillator frequency returns to normal. The illustration is a functional block diagram of a capacitive pressure sensor. The output of the sensor can be converted to digital data using an analog-to-digital converter (ADC) and then sent to a robot controller. Pressure sensors can be mounted in various places on a mobile robot, such as the front, back, and sides. Then, for example, physical pressure on the sensor in the front of the robot might send a signal to the controller, which tells the machine to move backward.  Capacitive Proximity Sensor Metal plates Dielectric foam Oscillator ADC Robot controller Capacitive pressure sensor A capacitive pressure sensor can be fooled by massive conducting or semiconducting objects in its vicinity. If such a mass comes near the transducer, the capacitance changes, even if direct contact is not made. This phenomenon is known as body capacitance. When the effect must be avoided, an elastomer can be used for pressure sensing. For proximity sensing, however, the phenomenon can be useful. See also CAPACITIVE PROXIMITY SENSOR, ELASTOMER, and PRESSURE SENSING. CAPACITIVE PROXIMITY SENSOR A capacitive proximity sensor takes advantage of the mutual capacitance that occurs between or among objects near each other. A capacitive proximity sensor uses a radio-frequency (RF) oscillator, a frequency detector, and a metal plate connected into the oscillator circuit, as shown in the diagram. The oscillator is designed so a change in the capacitance of the plate, with respect to the environment, causes the frequency to change. This change is sensed by the frequency detector, which sends a signal to the apparatus that controls the robot. In this way, if the system is properly designed, a robot can avoid bumping into things. In some detectors, the induced capacitance causes the oscillation to stop altogether.  Cartesian Coordinate Geometry Metal plate Oscillator Frequency detector ADC Sensed object Capacitive proximity sensor Robot controller Objects that conduct electricity to some extent, such as house wiring, people, cars, or refrigerators, are sensed more easily by capacitive transducers than are things that do not conduct, such as wooden chairs and doors. Therefore, other kinds of proximity sensors are necessary for a robot to navigate well in a complex environment, such as a household or office. Compare INDUCTIVE PROXIMITY SENSOR. See also PROXIMITY SENSING. CARTESIAN COORDINATE GEOMETRY Cartesian coordinate geometry is a common method by which a robot manipulator (arm) can move. This term derives from the Cartesian, or rectangular, coordinate system that is used for graphing mathematical functions. Alternatively, this movement scheme is called rectangular coordinate geometry. The drawing shows a Cartesian coordinate system in two dimensions. The axes are perpendicular to each other. In this case, they are up/down (vertical) and left/right (horizontal). Three-dimensional (3-D) Cartesian systems also exist. In a 3-D system, there are three linear axes, with each axis perpendicular to the other two. The manipulator shown in the illustration could be converted to 3-D Cartesian coordinate geometry by allowing the vertical rod to slide forward and backward (in and out of the page) along a horizontal track. Compare CYLINDRICAL COORDINATE GEOMETRY, POLAR COORDINATE GEOMETRY, REVOLUTE GEOMETRY, and SPHERICAL COORDINATE GEOMETRY.  Centralized Control Telescoping arm Sliding movement Cartesian coordinate geometry CENTRALIZED CONTROL In a system containing more than one robot, centralized control refers to oversight of all the individual robots by a single controller. Communication between the controller and the robots is usually done by wireless means such as radio, although other means, such as flexible wire or fiber-optic cables, can be used. This type of robotic system is somewhat analogous to a client-server computer network. In a centrally controlled robotic system, the main computer plays the role of a quasi-human operator. In some systems, the individual robots are partially autonomous, containing controllers of their own; this allows the system to keep operating at full capacity for a time, even in the event of a break in one or more of the communication links. This is known as partially centralized control. Another example of partially centralized control is a system in which each robot receives a set of instructions from  Charge-Coupled Device (CCD) the controller, stores those instructions, and then carries them out independently of the central controller. In some robotic systems, the individual units are completely and continuously dependent on the central controller, and cannot function if the communication link is severed. Such a system is said to employ fully centralized control. Compare DISTRIBUTED CONTROL. See also AUTONOMOUS ROBOT and INSECT ROBOT. CHAIN DRIVE A chain drive is a method of transferring mechanical energy in a robotic system from an actuator to a manipulator or end effector. It can also be used in wheel-drive propulsion systems. The system consists of a chain and a set of wheels with sprockets. The main asset of the chain drive is its simplicity. It can provide additional traction compared with a cable drive, because the chain is not likely to slip on the sprockets. Another asset is the fact that variable speed and power can be obtained by using sprockets of various sizes, in conjunction with a shifting mechanism. On the downside, the chain can come off the sprockets. The chain requires lubrication and maintenance, and can be noisy in operation. A common example of a chain drive is found in any bicycle. Compare CABLE DRIVE. CHARGE-COUPLED DEVICE (CCD) A charge-coupled device (CCD) is a camera that converts visible-light images into digital signals. Some CCDs also work with infrared (IR) or ultraviolet (UV). Common digital cameras work on a principle similar to that of the CCD. The image focused on the retina of the human eye, or on the film of a conventional camera, is an analog image. It can have infinitely many configurations, and infinitely many variations in hue, brightness, contrast, and saturation. A digital computer, however, needs a digital image to make sense of, and enhance, what it “sees.” Binary digital signals have only two possible states: high and low, or 1 and 0. It is possible to get an excellent approximation of an analog image in the form of high and low digital signals. This allows a computer program to process the image, bringing out details and features that would otherwise be impossible to detect. The illustration is a simplified block diagram of a CCD. The image falls on a matrix containing thousands or millions of tiny sensors. Each sensor produces one pixel (picture element). The computer (not shown) can employ all the tricks characteristic of any good graphics program. In  Charge-Coupled Device (CCD) Incoming light Scanning Signal processor Light-sensitive elements Hue control Brightness control Contrast control Enhanced video output Charge-coupled device addition to rendering high-contrast or false-color images, the CCD and computer together can detect and resolve images much fainter than is possible with conventional camera film or more primitive types of video cameras. This makes the CCD useful in robots that must employ night vision. Compare IMAGE ORTHICON and VIDICON. See also VISION SYSTEM.  Clean Room CHECKERS AND CHESS A computer can be programmed to play checkers. An excellent program was created by the roboticist Arthur Samuel, in which the computer can not only play the game move by move, but can also look ahead, or anticipate, to see the possible consequences of a move. Checkers is a fairly simple board game. It is more complex than tictac-toe, but far less sophisticated than chess. Anyone who has played tic-tac-toe has discovered that it is always possible to get at least a draw (tie). This is so elementary that a high-school student with some programming experience can get a computer to play tic-tac-toe. In this game the machine needs to look only one move ahead. Look-ahead strategy involving more than one move takes a certain amount of practice or learning. Computers can, however, be programmed to learn from their mistakes. Arthur Samuel’s checkers program uses a multiple-move look-ahead strategy so effectively that even the best human players in the world find it nigh impossible to beat his machine. There is another scheme that can be used for checkers: adopt a general game plan. General strategies can be broadly categorized as either defensive or offensive. The defensive/offensive schemes require look-ahead of only one move. Chess has been used to develop and test machine intelligence. One of the first chess-playing machines was developed by Rand Corporation in 1956. Chess is a complex game. A computer must look ahead more than one move to play a good game of chess. Multiple look-ahead strategy, along with machine learning, can enable a computer to play chess at a level of skill comparable to the masters. The program developed by Rand Corporation was able to prove some mathematical theorems. This is another good way to test the intelligence of a computer. CLEAN ROOM A clean room is a chamber specially designed and operated to minimize airborne contaminants. In some industries it is important that dust, dirt, bacteria, and other particulates be kept to an absolute minimum. A good example is the manufacture of integrated circuits (ICs) for electronic and computer systems. Robots have a considerable advantage over people in these environments. If certain precautions are observed, the environment in a room can be kept “clean” while allowing humans in. People who enter such a room must first put on airtight suits, gloves, and boots. A room that only robots enter, not people, can always be just a little bit cleaner.  Clinometer The contamination in a clean room is measured in terms of the number of particles of a certain size in 1 liter (1000 cubic centimeters) of air. Alternatively, the cubic foot is used as the standard unit of volume. See also INTEGRATED CIRCUIT. CLINOMETER A clinometer is a device for measuring the steepness of a sloping surface. Mobile robots use clinometers to avoid inclines that might cause them to tip too far, possibly even falling over. The floor in a building is almost always horizontal. Thus, its incline is zero. But sometimes there are inclines such as ramps. A good example is the kind of ramp used for wheelchairs, in which a very small elevation change occurs. A rolling robot cannot climb stairs, but it might use a wheelchair ramp, provided the ramp is not so steep that it upsets the robot’s balance or causes it to spill or drop its payload. A clinometer produces an electrical signal whenever it is tipped. The greater the angle of incline, the greater is the electrical output, as shown on the left side of the graph. A clinometer might also show whether an incline goes down or up. A downward slope might cause a negative voltage at the transducer output, and an upward slope a positive voltage, as shown on the right side of the graph. Positive output Output Down Up Slope Clinometer Negative output CLOSED-LOOP CONTROL Closed-loop control is a form of robot manipulator motion control in which the path, or trajectory, of the device is corrected at frequent intervals.  Coexistence After motion begins, a position sensor detects possible errors in the trajectory. If an error is detected, the sensor outputs a signal that operates through a feedback circuit to bring the manipulator back on course. The term derives from the fact that the feedback and control-signal circuits together constitute a closed loop. The main asset of closed-loop control is accuracy. In addition, closed-loop control can compensate for rapid, localized, or unexpected changes in the work environment. The principal disadvantages are greater cost and complexity than simpler schemes such as ballistic control. Compare BALLISTIC CONTROL. CLOSED-LOOP SYSTEM A closed-loop system is a set of devices that regulates its own behavior. Closed loops can be found in many kinds of machines, from the engine in a car (governor) to the gain control in a radio receiver (automatic level control). A closed-loop system, also known as a servomechanism, has some means of incorporating mechanical feedback from the output to the input. A sensor at the output end generates a signal that is sent back to the input to regulate the machine behavior. A good example of this is a back pressure sensor. Another example is closed-loop control of a robot manipulator. Compare OPEN-LOOP SYSTEM. See also BACK PRESSURE SENSOR, CLOSED-LOOP CONTROL, and SERVOMECHANISM. CMOS See COMPLEMENTARY METAL-OXIDE SEMICONDUCTOR. COEXISTENCE The term coexistence refers to programmed interactions among insect robots that share a working environment. The robots in such a system do not communicate directly with each other, but they all communicate with a central controller. There are three general schemes: ignorant coexistence, informed coexistence, and intelligent coexistence. In ignorant coexistence, none of the robots is aware that any of the others exists. In this sense, when two robots encounter one another, each machine regards its counterpart as an obstruction. Most mobile robots are programmed to avoid obstacles and hazards, maintaining a minimum distance of, say, 1 m. Thus, if there are numerous robots in a given environment and they all have ignorant coexistence, they tend to stay away from each other. If the robot “population density” is moderate to high, the machines tend to be more or less evenly spaced in the work environment at all times.  Cognitive Fatigue In informed coexistence, mobile robots can differentiate between obstructions or hazards and other robots. In this type of system, the robots are programmed to react or behave in a specific, but simple, way toward their counterparts. The most common behavior is for a robot to execute a specific set of movements when it senses the proximity of another robot, and a different set of movements when it senses the proximity of a nonrobotic obstruction or hazard. An example is for the machine to stop and reverse direction if it comes near an obstruction; but if it comes near another robot, it stops, waits a second, and if the other robot remains in the way, turns right 90°, proceeds 1 m, then turns left 90° and resumes moving in the original direction. In intelligent coexistence, as in informed coexistence, the robots can differentiate between obstructions or hazards and other robots. However, the programmed response is more sophisticated. For example, each robot might be programmed to avoid coming within 1 m of any other robot. If such an approach does occur, triggering the avoidance response, the robot is programmed to move in a direction corresponding to the average direction of all the other robots in the system. Each robot obtains this general information from the controller. Compare COOPERATION. See also AUTONOMOUS ROBOT, CENTRALIZED CONTROL, DISTRIBUTED CONTROL, and INSECT ROBOT. COGNITIVE FATIGUE Cognitive fatigue is a form of mental exhaustion sometimes experienced by users of telepresence systems. Most teleoperated systems must compromise realism in order to keep within limitations imposed by available bandwidth and allowable expense. In a typical telepresence system, the cameras usually lack peripheral vision. Signal propagation delays can cause latency problems (time lag between command and response), particularly when teleoperation is done over long distances. Image resolution (detail) and refresh rate (the number of video frames per second) are generally compromised. Audio systems are generally better than video systems because the necessary bandwidths are smaller, but tactile sensation is poor or absent. Symptoms of cognitive fatigue include wandering attention, sleepiness, headache, and irritability. These problems can result in equipment operation errors. See also TELEPRESENCE. COGNIZANT FAILURE Cognizant failure is a feature of machine intelligence in which a failed subsystem or program is replaced by one at a higher level, while ensuring  Color Sensing that all processes continue to run smoothly without undesired side effects. In uncomplicated systems, a high-level part of the system can temporarily take over the tasks of a lower-level part, without regard for details of the event. In scenarios where the possibilities are diverse and variable, certain additional procedural steps, not normally necessary, are sometimes required to ensure smooth operation while the lower-level device or subsystem is repaired. Consider the case of a smart home equipped with smoke detectors, heat sensors, a telephone link to the fire department, and a set of sprinklers. What should the system do if a heat sensor is set off by a mischievous child with a hair dryer, causing a false alarm? An unsophisticated system calls the fire department and actuates the sprinklers, causing embarrassment and unnecessary damage to furniture. A sophisticated system can prevent these undesirable things from taking place, provided the owner of the house, or some backup sensing system, is present to determine that there is actually no fire. The owner or backup system must be cognizant of the fact that the alarm is false. Then the sprinkler system can be disabled, a call can be made to the fire department to cancel the alarm, and the offending sensor, if it has been damaged, can be shut down until it is replaced. (The child can be disciplined as well, although this is the responsibility of the human home owner.) Compare GRACEFUL DEGRADATION. COLOR SENSING Many robotic vision systems function only in grayscale. Color sensing can be added, in a manner similar to the way it is added to television (TV) systems. Color sensing can help a robot determine the identity or nature of an object. Is an observed horizontal surface a floor or a grassy yard? (If it is green, it is probably a grassy yard.) Sometimes, objects have regions of different colors that have identical brightness as seen by a grayscale system. Such objects can be analyzed better with a color-sensing system than with a vision system that sees only shades of gray. The drawing shows a block diagram of a color-sensing system. Three grayscale cameras are used. Each camera has a color filter in its lens. One filter is red, another is green, and another is blue. These are the three primary colors of radiant light. All possible hues, brightnesses, and saturations are comprised of these three colors in various ratios. The signals from the three cameras are processed by a microcomputer, and the result is fed to the robot controller. See also GRAYSCALE, OBJECT RECOGNITION, TEXTURE SENSING, and VISION SYSTEM.  Competing Sensors Red Camera Illumination Green Camera from scene Blue Camera Image processor Color filters Color sensing Robot controller COMPETING SENSORS See SENSOR COMPETITION. COMPLEMENTARY METAL-OXIDE SEMICONDUCTOR (CMOS) Complementary metal-oxide semiconductor, also called CMOS (pronounced “seamoss”), is the name for a technology used in digital devices, such as computers. Two types of field-effect transistor (FET) work together, in tandem and in huge numbers, on a single integrated circuit (IC) chip. The main asset of CMOS technology in robotics is the fact that the devices can function effectively with tiny electrical currents. Thus, wellengineered CMOS circuits draw very little power from the power supply, allowing the use of batteries. Another advantage of CMOS technology is that it works extremely fast. It can process a lot of data in a short period of time. A disadvantage of CMOS devices is the fact that they are easily damaged by static electricity. Devices of this type must be stored with their pins embedded in conductive foam material, and/or packaged in special plastic that resists electrostatic-charge buildup. When constructing or servicing equipment using CMOS, technicians must take precautions to avoid the presence of static electric charges on their hands, and on instruments  Compliance such as probes and soldering irons. This is usually ensured by physically connecting the technician’s body to a good electrical ground. See also INTEGRATED CIRCUIT. COMPLEX-MOTION PROGRAMMING As machines become smarter, the programming gets more sophisticated. No machine has yet been built that has intelligence anywhere near that of a human being. Some researchers think that true artificial intelligence (AI), at a level near that of the human brain, will never be achieved. The programming of robots can be divided into levels, starting with the least sophisticated and progressing to the theoretical level of true AI. The drawing shows a four-level scheme. Level 2, just below the task level but above the simple-motion level, is called complex-motion programming. Robots at this level can perform sets of motions in defined sequences. Compare ARTIFICIAL INTELLIGENCE, SIMPLE-MOTION PROGRAMMING, and TASK-LEVEL PROGRAMMING. Fourth level Third level Second level First level Artificial intelligence Tasks Complex motions Simple motions Complex-motion programming COMPLIANCE Compliance is the extent to which a robot end effector or manipulator moves, or yields, when a force is applied to it. It can be expressed qualitatively (using terms such as “springy” or “rigid”) or quantitatively in terms of displacement per unit force (such as millimeters per newton). A robot is said to be compliant if its mechanical movements are affected by external forces, including linear pressure or torque. The compliance can occur along one, two, or three axes, or in a rotational sense. Generally, a compliant robot should be adjusted so the behavior of its  Composite Video Signal manipulators and end effectors keeps the stress on its components to a minimum. One means of accomplishing this is a back pressure sensor. See also BACK PRESSURE SENSOR. COMPOSITE VIDEO SIGNAL A composite video signal is the waveform that modulates a television (TV) or video carrier. The composite signal contains video intelligence as well as synchronization, blanking, and timing pulses. The bandwidth is typically 6 MHz (6 megahertz) for conventional fast-scan signals, and approximately 3 kHz (3 kilohertz) for slow-scan signals. A video camera, such as an image orthicon or vidicon, produces a fast-scan signal. Some robotic vision systems generate and analyze composite video signals. The illustration shows the waveform for a single line of a color picture signal. There are normally 525 or 625 lines in a complete frame for standard Blanking pulses Amplitude Image data Black White Time 63.5 microseconds Composite video signal  Configuration Space (C-Space) fast-scan video. In robotic vision systems, there are advantages to using more lines per frame than is standard with television, in order to obtain improved image resolution. See also IMAGE ORTHICON, VIDICON, and VISION SYSTEM. COMPUTER MAP An autonomous robot must have a sense of where it is relative to surrounding objects, so that it will not bump into things, and so that it can find whatever it is seeking. For this to be possible, the robot controller can make a computer map of its environment. Computer maps can be generated using radar, sonar, or a vision system. Such a map can exist in either two or three dimensions. A two-dimensional (2-D) computer map of the objects in a room might be generated for a flat plane 1 m above the floor. Several 2-D maps, representing various altitudes above the floor, can be combined to create a composite threedimensional (3-D) map. A more sophisticated method of generating a 3-D computer map involves the use of spherical coordinates. The spherical coordinate system defines azimuth (compass bearing), elevation (angle above the horizontal), and range (radial distance). For such a map to serve its purpose, hundreds or even thousands of individual soundings or observations must be made. These soundings or observations should be distributed evenly around a half-sphere above the horizontal for terrestrial robots, or around a full sphere for submarine, airborne, or deep-space robots. In deep space, a reference plane must be chosen to serve as the “horizontal.” The larger the number of soundings, the better is the resolution of the map. See also RADAR, SONAR, and VISION SYSTEM. CONFIGURATION SPACE (C-SPACE) A configuration space (abbreviated C-space) is a scheme in which the location and orientation of a robot is determined relative to other objects in its environment. Ideally, a C-space should use the minimum number of coordinates necessary to accomplish this task. This eliminates redundancy, which consumes controller memory and can cause confusion. Consider a mobile robot designed to function on a single floor of a building. The total physical region in which this robot exists (the world space) is three-dimensional (3-D). This constitutes three degrees of freedom, which can be considered in terms of the Cartesian (rectangular) coordinates x (north/south), y (east/west), and z (up/down). The orientation, or attitude, of the robot, can require up to three additional degrees of freedom: p (pitch), r (roll), and w (yaw).  Contact Sensor On the flat plane of a floor, the location of the robot can be denoted in two-dimensional (2-D) coordinates. In the Cartesian system described above, these are x and y. If the attitude nevertheless requires that p, r, and w each be specified, the C-space requires five degrees of freedom: x, y, p, r, and w. However, p, r, and w may not all be important in the 2-D case. This could reduce the number of degrees of freedom in the C-space still further. Compare WORK ENVELOPE and WORK ENVIRONMENT. See also CARTESIAN COORDINATE GEOMETRY, PITCH, ROLL, and YAW. CONTACT SENSOR A contact sensor is a device that detects objects, obstructions, or barriers by means of direct physical contact. Contact sensors can also be used to measure applied force or torque. In robotics, such devices include “whiskers” and pressure sensors. Simplicity is the main asset of contact sensing when used to determine the presence or absence of an object. In order to measure force or torque accurately, especially when such force or torque must be regulated, a closed-loop system is required. See also CLOSED-LOOP CONTROL, CLOSED-LOOP SYSTEM, FEEDBACK, and PRESSURE SENSING. CONTEXT Context is the environment in which a word is used. It is important in speech recognition systems, such as those used in personal or security robots designed to respond to spoken commands. Everyone has heard the expression “out of context.” When a word is used out of context, it results in a phrase or sentence that does not make sense. Worse yet, it might mean something not intended. When a word is taken out of context, the phrase or sentence is technically all right, but it is interpreted as nonsense, or in the wrong way. In order to interpret and respond to spoken statements properly, a computer or robot with artificial intelligence must know the context in which each word is used. Humans have an innate sense of context; machines do not. This makes the design and programming of effective speech recognition systems an extremely sophisticated business. See also PROSODIC FEATURES and SPEECH RECOGNITION. CONTINUOUS ASSISTANCE See SHARED CONTROL. CONTINUOUS-PATH MOTION A robot arm can move smoothly or in discrete steps. Smooth-moving robot manipulators employ continuous-path motion.  Controller For a robot to move along a smooth, continuous path, every point along the way must, in theory, be stored in the controller memory. Of course, this is not literally possible, because a continuous path contains an infinite number of points. Continuous-path motion uses mathematical functions, rather than point sets, to define the instantaneous position of a robot manipulator. In the function method, the instantaneous position is stored as a set of mathematical functions. Such motion is truly continuous, in that it actually passes through an infinite number of points. This is possible because of the smooth nature of the mathematical functions. This principle is the robot-motion analog of vector graphics in computing. Compare POINT-TOPOINT MOTION. CONTROLLER In a robot, the controller is a computer that oversees and controls the operation and motion of the machine. The illustration is a functional block diagram of a controller. The heart of the controller is the central processing unit (CPU), which is similar to the CPU in a personal computer. Movement instructions are held in random-access memory (RAM) and/or on storage media such as a hard drive. The interface does several things. Mainly, it allows the microcomputer to communicate with a human operator or supervisor. Through the interface, To human operator Interface Memory CPU To robot manipulator(s) Controller  Control Trading it is possible to reprogram the memory to change the movement instructions. The actions or function repertoire of the robot can be displayed on a monitor screen. There might also be various malfunction indicators. Some of the more sophisticated interfaces have a teach box, which lets the human operator reprogram the motions and path of the robot. See also TEACH BOX. CONTROL TRADING Control trading is a limited form of robotic remote control in a system that employs teleoperation. The operator instructs the robot to perform a specific, complete task, such as vacuuming a room or mowing a lawn. The machine then carries out the entire task without further instruction or supervision by the human. Control trading has obvious assets. The human operator does not have to constantly monitor the progress of the machine, although periodic checking is advisable to ensure that a major malfunction does not occur. It is thus possible for a single operator to oversee the operation of several robots at the same time. Another asset is the fact that latency, or the time lag caused by signal propagation delays, is not a serious problem. Control trading is ideal, for example, in the teleoperation of a robot on Mars, or the teleoperation of an interplanetary space probe. Still another asset is that large signal bandwidth is not required, especially for the uplink to the machine; commands can consist of encoded messages of a relatively small number of bytes. The main limitation of control trading is the fact that the robot cannot be expected to contend with sudden, unforeseen changes in the work environment. The machine performs its programmed set of operations under the assumption that the environment will cooperate. In scenarios where the robot work environment is subject to frequent change, shared control is generally superior to control trading. Compare SHARED CONTROL. See also TELEOPERATION. COOPERATION Cooperation is constructive or synergistic interaction of robots in a system. It can take various forms, depending on the manner and extent to which the robots communicate, and the degree of autonomy each machine has. In nonactive cooperation, the robots do not necessarily have to communicate. However, it is important that each robot be able to tell the other robots apart from general objects in the environment. This prevents undesirable conditions such as collisions between robots, multiple robots attempting the same task at the same time and in the same place, and uneven distribution of the machines in the work environment. Other than the ability to avoid conflicting with its peers, each robot in a  Correspondence non-active-cooperative system need not pay particular attention to the others. In a well-designed system of this kind, cooperation occurs naturally. In active cooperation, the robots are capable of acknowledging one another, and in some cases communicating with and assisting each other as well. Active cooperation can range from “loose,” in which the machines are aware of each other’s existence and function but do not communicate, to “tight,” in which each robot can communicate with any or all of the others. Some systems can be engineered to exhibit cooperative mobility, in which two or more robots can combine in “special teams” to deal with complex or difficult tasks that a single robot cannot carry out. A special form of active cooperation involves centralized control, in which the robots are all dependent on oversight by a single controller. Compare COEXISTENCE. See also AUTONOMOUS ROBOT, CENTRALIZED CONTROL, DISTRIBUTED CONTROL, and INSECT ROBOT. CORRESPONDENCE In binocular machine vision, the term correspondence refers to the focusing of both video cameras or receptors on the same point in space. This ensures that the video perception is correct. If the two “eyes” are not focused at the same point, the ability of the machine to perceive depth is impaired. The human sense of correspondence can be confused when looking at a grid of dots, or at a piece of quadrille graph paper. The illustration shows Eyes on same object Eyes not on same object Eyes not on same object Focal distance correct Correspondence Focal distance too long Focal distance too short  Cryptanalysis two ways that human eyes, or a machine vision system, can be fooled by such a pattern. This problem is generally limited to observations of regular patterns of dots, squares, or other identical objects. It rarely occurs in complex scenes in which geometric shapes do not repeat. On the left in the illustration, both video sensors (shown as eyeballs) are looking at the same point. Thus, depth is perceived correctly, even if the views of the object appear slightly different because of the difference in viewing angle through either sensor. In the drawings in the center and on the right, the left sensor is looking at one object in the set, while the right sensor looks at another. Because all the objects are evenly spaced, they seem to line up as perceived by the vision system. If a robot manipulator acts on this incorrect information, positioning errors are likely. See also BINOCULAR MACHINE VISION. CRYPTANALYSIS Cryptanalysis is the art of breaking ciphers, which are signal-processing schemes used to keep unauthorized people from intercepting communications or gaining access to sensitive data. With the help of computers, cryptanalysis has become much more sophisticated than it once was. A computer can test different solutions to a code much more rapidly than teams of humans ever could. Beyond that, artificial intelligence (AI) can be employed in an attempt to figure out what an enemy is thinking. This streamlines the process of cipher breaking. It lets the cryptanalyst, or code breaker, get a feel for the general scheme behind a cipher, and in this way, it helps the cryptanalyst understand the subtleties of the code more quickly. One of the earliest cryptanalysts to use a computer was Alan Turing, known as a pioneer in AI. In the early 1940s, during World War II, the Germans developed a sophisticated machine called Enigma that encoded military signals. The machine and its codes confounded Allied cryptanalysts, until Alan Turing designed one of the first true computers to decode the signals. As computers become more powerful, they can create more complex ciphers. But they can also invent increasingly sophisticated decryption schemes. In warfare, the advantage in encryption/decryption goes to the side with the more advanced AI technology. CYBERNETICS The term cybernetics refers to the science of goal-seeking, or self-regulating, things. The word itself comes from the Greek word for “governor.” The fields of robotics and artificial intelligence are subspecialties within  Cylindrical Coordinate Geometry the science of cybernetics. Computer-controlled robots that interact with their environments are cybernetic machines. An example of a cybernetic process is pouring a cup of coffee. Suppose someone says to a personal robot, “Please bring me a cup of coffee, and be sure it’s hot.” In the robot controller’s memory, there are data concerning what a coffee cup looks like, the route to the kitchen, the shape of the coffee pot, and a relative-temperature-interpretation routine, so the robot knows what the person means by “hot.” A personal robot must go through an unbelievably complicated process to get a cup of coffee. This becomes evident when one tries to write down each step in rigorous form. CYBORG The word cyborg is a contraction of “cybernetic” and “organism.” In robotics, the term refers to a human whose body is comprised largely, or even mostly, of robotic elements, but who is still biologically alive. If a person is given a single robotic hand or arm, it is called a bionic body part or prosthesis. Science fiction carries this notion to the point that a person seriously injured might be reconstructed significantly, or even almost entirely, of bionic parts. Such a being would be a true cyborg. Technology is a long way from creating cyborgs, but some scientists believe they will someday be common. A few futurists envision a society comprised of human beings, cyborgs, smart robots, and computers. This has been called a cybot society. While enthusiasm for the idea of a cybot society runs high in Japan, there is somewhat less interest in the United States and Europe. Americans and Europeans think of robots as serving mainly industrial purposes, but the Japanese think of them as being in some sense alive. This might be why the Japanese are so much more active in developing humanlike robots. See also ANDROID and PROSTHESIS. CYLINDRICAL COORDINATE GEOMETRY Cylindrical coordinate geometry, also known as cyclic coordinate geometry, is a scheme for guiding a robot arm in three dimensions. A cylindrical coordinate system is a polar system with an extra coordinate added for elevation. Using this system, the position of a point can be uniquely determined in three-dimensional (3-D) space. In the cylindrical system, a reference plane is used. An origin point is chosen, and also a reference axis, running away from the origin in the reference plane. In the reference plane, the position of any point can be specified in terms of the reach, or distance from the origin, and the base rotation, which is the angle measured counterclockwise from the reference axis.  Cylindrical Coordinate Geometry The elevation coordinate is either positive (above the reference plane), negative (below it), or zero (in it). The illustration shows a robot arm equipped for cylindrical coordinate geometry. Compare CARTESIAN COORDINATE GEOMETRY, POLAR COORDINATE GEOMETRY, REVOLUTE GEOMETRY, and SPHERICAL COORDINATE GEOMETRY. Sliding and rotating movement Telescoping arm Reference plane Reference axis Cylindrical coordinate geometry  D DATA COMPRESSION Data compression is a method of maximizing the amount of digital information that can be stored in a given space, or sent in a certain period of time. Text and program files can be compressed by replacing often-used words and phrases with symbols such as =, #, &, or @, as long as none of these symbols occurs in the uncompressed file. When the data are received, they are uncompressed by substituting the original words and phrases for the symbols. Digital images can be compressed in either of two ways. In lossless image compression, detail is not sacrificed; only the redundant bits are eliminated. In lossy image compression, some detail is lost, although the loss is usually not significant. Text and programs can generally be reduced in size by about 50 percent by means of data compression. Images can be reduced to a much larger extent if a certain amount of loss can be tolerated. Some advanced imagecompression schemes can output a file that is only a tiny fraction of the original file size. DATA CONVERSION Many communications systems “digitize” analog signals at the source (transmitting end) and “undigitize” the signals at the destination (receiving end). Digital data can be transferred bit by bit (serial) or in bunches (parallel). Data conversion is the process of altering data between analog and digital forms, or between parallel and serial forms.  Data Conversion Analog to digital Any analog, or continuously variable, signal can be converted into a string of pulses whose amplitudes have a finite number of states. This is analog-to-digital (A/D) conversion. An A/D converter or ADC samples the instantaneous amplitude of an analog signal and outputs pulses having discrete levels, as shown in Fig. 1. The number of levels is called the sampling resolution, and is usually a power of 2. The number of pulses per second is the sampling rate. The time between pulses is the sampling interval. In this example, there are eight levels, represented by three-digit binary numbers from 000 to 111. 111 110 101 100 011 010 001 000 Time Sampling interval Analog waveform Data conversion—Fig. 1 In general, the minimum workable digital sampling rate is approximately twice the highest analog data frequency. This is a general principle in communications engineering, known as the Nyquist theorem or sampling theorem. For a signal with components as high as 3 kHz, the minimum sampling rate is 6 kHz. The commercial voice standard is 8 kHz. For hi-fi music transmission, the standard sampling rate is 44.1 kHz. In machine communications systems, the minimum sampling rate depends on the speed with which data must be transferred between points, such as from a central controller to a mobile robot.  Data Conversion Digital to analog The scheme for digital-to-analog (D/A) conversion depends on whether the signal is binary or multilevel. The D/A conversion process is carried out by a D/A converter (DAC). In a binary DAC, a microprocessor reverses the A/D conversion process done in recording or transmission. Multilevel digital signals can be converted back to analog form by “smoothing out” the pulses. This can be intuitively seen by examining Fig. 1. Imagine the train of pulses being smoothed into the continuous curve. Digital signals lend themselves to repeated reproduction without loss of integrity. Digital signals are also relatively immune to the effects of noise in wireless and long-distance cable circuits. For this reason, even if the initial input and final output signals are analog in nature, such as moving images or human voices, there are advantages to using digital format in the intervening medium. Digital signals can be clarified by means of digital signal processing (DSP) to enhance the signal-to-noise (S/N) ratio, thereby minimizing the number of communication errors and necessary bandwidth while maximizing the data transfer rate. This is true whether the ultimate input and output signals are analog or digital. Serial versus parallel Binary data can be sent and received one bit at a time along a single line or channel. This is serial data transmission. Higher data speeds can be obtained by using multiple lines or a wideband channel, sending independent sequences of bits (high and low, or 1 and 0) along each line or subchannel. This is parallel data transmission. In parallel-to-serial (P/S) conversion, bits are received from multiple lines or channels, and transmitted one at a time along a single line or channel. A buffer stores the bits from the parallel lines or channels while they are awaiting transmission along the serial line or channel. In serial-to-parallel (S/P) conversion, bits are received from a serial line or channel, and sent in batches along several lines or channels. The output of an S/P converter cannot go any faster than the input, but the circuit is useful when it is necessary to interface between a serial-data device and a parallel-data device. Figure 2 illustrates a communications circuit in which a P/S converter is used at the source and an S/P converter is used at the destination. In this example, the data characters are 8-bit bytes; the illustration shows the transfer of one character.  Decimal Number System Source 1 0 1 0 0 1 1 0 P/S 1 0 1 Data flow 0 0 1 1 0 S/P 1 0 1 0 0 1 1 0 Destination Data conversion—Fig. 2 DECIMAL NUMBER SYSTEM See NUMERATION. DEGREES OF FREEDOM The term degrees of freedom refers to the number of different ways in which a robot arm can move. Most robot arms move in three dimensions, but they often have more than three degrees of freedom.  Degrees of Rotation You can use your own arm to get an idea of the degrees of freedom that a robot arm might have. Extend your right arm straight out toward the horizon. Extend your index finger so it is pointing. Keeping your arm straight, move it from the shoulder. You can move your arm three ways. Up-anddown movement is called pitch. Movement to the right and left is yaw. You can also rotate your whole arm as if you were using it as a screwdriver; this is roll. Your shoulder has three degrees of freedom: pitch, yaw, and roll. Now move your arm from the elbow only. If you hold your shoulder and upper arm in the same position constantly, you can see that your elbow joint has the equivalent of pitch in your shoulder joint. But that is all (unless your elbow is dislocated). The human elbow has one degree of freedom. Extend your arm toward the horizon, straighten it out, and move only your wrist. Keep the arm above the wrist straight and motionless. Your wrist can bend up and down and can also move side to side. The human hand has two degrees of freedom with respect to the arm above it: pitch and yaw. Thus, in total, your shoulder/elbow/wrist system has six degrees of freedom: three in the shoulder, one in the elbow, and two in the wrist. A certain amount of roll is also possible in the arm below the elbow; this does not occur in either the elbow joint or the wrist joint, but in the lower arm itself. This makes for a seventh degree of freedom. Three degrees of freedom are sufficient to bring the end of a robot arm to any point within its work envelope, or work space, in three dimensions. Thus, in theory, it might seem that a robot should never need more than three degrees of freedom. But the extra possible motions, provided by multiple joints, give robot arms versatility that they could not have with only three degrees of freedom. See also ARTICULATED GEOMETRY, CARTESIAN COORDINATE GEOMETRY, CYLINDRICAL COORDINATE GEOMETRY, DEGREES OF ROTATION, POLAR COORDINATE GEOMETRY, REVOLUTE GEOMETRY, ROBOT ARM, SPHERICAL COORDINATE GEOMETRY, and WORK ENVELOPE. DEGREES OF ROTATION Degrees of rotation are a measure of the extent to which a robot joint, or a set of robot joints, is turned. Some reference point is always used, and the angles are specified in degrees or radians with respect to that joint. Rotation in one direction (usually clockwise) is represented by positive angles; rotation in the opposite direction is specified by negative angles. Thus, if angle X = 58°, it refers to a rotation of 58° clockwise with respect to the reference axis. If angle Y = 74°, it refers to a rotation of 74° counterclockwise. The illustration shows a robot arm with three joints. The reference axes are J1, J2, and J3, for rotation angles X, Y, and Z. The individual angles add together.  Deliberation J3 Z = 51 degrees Y = 75 degrees J2 J1 X= 39 degrees X + Y + Z = 165 degrees Degrees of rotation When it is necessary to move this robot arm to a certain position within its work envelope, or the region in space that the arm can reach, the operator enters data into a computer. This data includes the measures of angles X, Y, and Z. In the example shown by the illustration, the operator has specified X = 39°, Y = 75°, and Z = 51°. For simplicity, no other possible variable parameters, such as base rotation, wrist rotation, or extension/retraction of linear sections, are shown. See also ARTICULATED GEOMETRY, DEGREES OF FREEDOM, ROBOT ARM, and WORK ENVELOPE. DELIBERATION Deliberation refers to any characteristic of robotic navigation that involves advance planning of some sort, rather than mere reaction to the presence  Depth Map of obstacles or changes in the work environment. Deliberative planning is commonly combined with another scheme called reactive planning. See also HIERARCHICAL PARADIGM, HYBRID DELIBERATIVE/REACTIVE PARADIGM, and REACTIVE PARADIGM. DEPTH MAP A depth map, also called a range image, is a specialized form of computer map, rendered as a grayscale image of a robot’s work environment. The brightness of each pixel (picture element) in the image is proportional to the range, or radial distance, to the nearest obstruction in a specific direction. In some depth maps, the brightest pixels correspond to short range; in others, the brightest pixels correspond to long range. A typical range image looks something like a grayscale video image or its negative. However, upon examination, the difference between a conventional visible or infrared (IR) image and a depth map becomes apparent. Local detail in objects, such as the contour of a human face, generally do not show up in a depth map, even if the shade, color, or heat radiation vary greatly. It is the radial distance, as determined by a range sensing and plotting system, which produces the image. Suppose a robot is navigating across a flat field or empty parking lot on which a huge ball sits. The range sensing and plotting system is programmed to produce a depth map. In the field of view of the system, the only objects that appear are the flat surface and the ball. Suppose the depth map is such that the relative brightness of the image is inversely proportional to the radial distance. The depth map looks like the rendition shown in the accompanying illustration. The color of the ball and the surface on Depth map  Derivative which it rests, and the time of day or night, do not matter; the rendition is based entirely on the range as a function of the direction in threedimensional space. See also COMPUTER MAP and RANGE SENSING AND PLOTTING. DERIVATIVE The term derivative refers to the rate of change of a mathematical function. For example, speed or velocity is the derivative of displacement, and acceleration is the derivative of velocity. Figure 1 shows a hypothetical graph of displacement as a function of time. This function appears as a curve. You might think of it as a graph of the distance traveled by a robot accelerating along a linear track, with the displacement specified in meters and the time in seconds. At any specific instant in time, call it t, the speed is equal to the slope of the line tangent to the curve at that moment. This quantity is expressed in linear displacement units (such as meters) per second. Displacement Position vs. Time Slope = speed at time t Time t Derivative—Fig. 1 In digital electronics, a circuit that continuously takes the derivative of an input wave, as a function of instantaneous amplitude versus time, is called a differentiator. An example of the operation of a differentiator is shown in Fig. 2. The input is a sine wave. The output follows the slope, or derivative, of this wave; the result is a cosine wave, with the same shape as the sine wave but displaced by one-fourth of a cycle (90° of phase). Compare INTEGRAL.  Differential Ampifier Amplitude Function + Time _ Derivative Derivative—Fig. 2 DICHOTOMIZING SEARCH See BINARY SEARCH. DIFFERENTIAL AMPLIFIER A differential amplifier is an electronic circuit that responds to the difference in amplitude between two signals. Some differential amplifiers also produce gain, resulting in an output signal whose amplitude varies dramatically when the amplitude of either input signal varies only a little. The output is proportional to the difference between the input signal levels. If the input amplitudes are identical, then the output is zero. The nomograph shows how the instantaneous output of a differential amplifier varies as the instantaneous input values change. To find the Differential amplifier  Differential Transducer output, place a straight ruler so its edge passes through the two input points; the output is the point on the center scale through which the ruler passes. In this example, the circuit has no gain. Differential amplifiers are sometimes employed in robotic sensing systems. The output of an amplifier in this situation can be used as an error signal, which is sent to the guidance system to regulate the movement of a mobile robot. This can ensure that the robot follows a prescribed route in its work environment, such as the path along which two reference acoustic or radio waves are exactly in phase. Compare DIFFERENTIAL TRANSDUCER. DIFFERENTIAL TRANSDUCER A differential transducer is a sensing device with two inputs and one output. The output is proportional to the difference between the input signal levels. An example is a differential pressure transducer, which responds to the difference in mechanical pressure at two points. Any pair of transducers can be connected in a differential arrangement. Usually, this involves connecting the transducers to the inputs of a differential amplifier. When the two variables have the same magnitude, the output of the differential transducer is zero. The greater the difference in the magnitudes of the sensed effects, the greater is the output. The most output occurs when one of the sensed effects is intense, and the other is zero or near zero. Whether the output is positive or negative depends on which of the sensed effects is greater. Compare DIFFERENTIAL AMPLIFIER. DIFFERENTIATION See DERIVATIVE. DIGITAL IMAGE A digital image, also called a digitized image, is a rendition of a scene at visible, infrared (IR), or ultraviolet (UV) wavelengths, or using radar or sonar, in the form of a rectangular array of tiny squares or dots called pixels. In a grayscale digital image, each pixel has a brightness level that can attain any of numerous discrete binary values. Common ranges are from binary 0000 through 1111 (16 shades of gray) or binary 00000000 through 11111111 (256 shades of gray). In a color digital image, each pixel has a color value of red, green, or blue (RGB), and also a brightness level that can attain any of numerous discrete binary values. Color digital images occupy considerably more data memory or storage space than grayscale digital images, because the three color values can vary independently with each pixel. The number of pixels in a digital image determines the resolution. This figure is generally represented in terms of the number of pixels in  Digital Signal Processing (DSP) the horizontal and vertical dimensions. In a computer display, for example, a common resolution is 1024 768 (1024 pixels horizontal and 768 pixels vertical). In a visible digital image, color is usually rendered in as true-to-life a fashion as possible. However, at IR and UV wavelengths, and especially with radar and sonar, false colors are often used in digital images. For example, in a sonar image, color can represent the range, or the distance between a robot and objects in its work environment. Red might represent the smallest range, progressing up through orange, yellow, green, blue, violet, and finally white, representing the greatest (or infinite) range. See also RESOLUTION. DIGITAL INTEGRATED CIRCUIT See INTEGRATED CIRCUIT. DIGITAL LOGIC See LOGIC GATE. DIGITAL MOTION In robotics, digital motion refers to the movement of a robot arm that can stop only at certain positions within its work envelope. This is in contrast to analogical motion, in which the number of possible positions is theoretically infinite. The possible positions in a system that incorporates digital motion must be programmed into the robot controller. For example, the base of a robot arm might rotate to any multiple of 30° in the complete circle of 0° to 360°. This allows for 12 unique base-rotation angles. If more precision is required, the angle increment can be reduced (10° will allow for 36 unique base-rotation angles, for example). When the robot arm must be rotated to a certain base angle position, the desired angle or step is entered into the robot controller. The arm then moves to the designated position and stops. Stepper motors are commonly used in robotic digital-motion systems. These motors move in discrete increments, rather than rotating continuously. Compare ANALOGICAL MOTION. See also STEPPER MOTOR. DIGITAL SIGNAL PROCESSING (DSP) Digital signal processing (DSP) is a scheme for improving the precision of digital data. It can be used to clarify or enhance signals of all kinds. Analog cleanup When DSP is used in an analog communications system, the signal is first changed into digital form by A/D conversion. Then the digital signal is  Digital Signal Processing (DSP) “tidied up” so the pulse timing and amplitude adhere strictly to protocol. Finally, the digital signal is changed back to analog form by means of D/A conversion. Digital signal processing can extend the workable range of a communications circuit, because it allows reception under worse conditions than would be possible without it. Digital signal processing also improves the quality of fair signals, so the receiving equipment or operator makes fewer errors. The DSP process also ensures that the necessary communications bandwidth is kept to a minimum. Digital cleanup In circuits that use only digital modes, A/D and D/A conversion are irrelevant, but DSP can nevertheless “tidy up” the signal. This improves the accuracy of the system, and also makes it possible to copy data many times (that is, to produce multigeneration copies). The DSP circuit minimizes confusion between digital states, as shown in the illustration. A hypothetical signal before processing is shown at the top; the signal after processing is shown at the bottom. If the input amplitude is above a certain level for an interval of time, the output is high (logic 1). If the input amplitude is below the critical point for a time interval, then the output is low (logic 0). A strong burst of noise might Relative amplitude Signal before DSP Time Signal after DSP Digital signal processing  Direction Finding fool the circuit into thinking the signal is high when it is actually low; but overall, errors are less frequent with DSP than without it. In computers and robots A DSP system can be etched onto a single integrated circuit (IC), similar in size to a memory chip. Some DSP circuits serve multiple functions in a computer or robotic system, so the controller can devote itself to doing its primary work without having to bother with extraneous tasks. A DSP chip can compress and decompress data, help a computer recognize and generate speech, translate from one spoken language to another (such as from English to Chinese or vice versa), and recognize and compare patterns. See also DATA CONVERSION. DIRECTIONAL TRANSDUCER A directional transducer is a device that senses some effect or disturbance, and produces an output signal that varies in amplitude depending on the direction from which the effect or disturbance arrives. Directional transducers are used extensively in robotic sensing and guidance systems. A simple example of a directional transducer is a common microphone. Microphones are almost always unidirectional, that is, they respond best in one direction. An example of a bidirectional transducer is a horizontal radio antenna known as a dipole. Some transducers are omnidirectional in a specified plane. A vertical radio antenna is an example. It works equally well in all horizontal directions. However, its sensitivity varies in vertical planes. Some transducers are equally sensitive in all possible directions; the directional pattern for such a device is a sphere in three dimensions. This is a truly omnidirectional transducer. DIRECTION FINDING Direction finding is a means of location and/or navigation, usually employing radio or acoustic waves. At radio frequencies (RF), location and navigation systems operate between a few kilohertz and the microwave region. Acoustic systems use frequencies between a few hundred hertz and a few hundred kilohertz. Signal comparison A mobile robot can find its position by comparing the signals from two fixed stations whose positions are known, as shown in Fig. 1. By adding 180° to the bearings of the sources X and Y, the robot (square) obtains its bearings as “seen” from the sources (dots). The robot can determine its direction and speed by taking two readings separated by a certain amount of time. Computers can assist in precisely determining, and displaying, the position and the velocity vector.  Direction Finding Source Y Source X Robot Direction finding—Fig. 1 Ultrasonic transducer Amplifier Servo Amplitude Meter To robot controller Direction finding—Fig. 2 Figure 2 is a block diagram of an acoustic direction finder. In this case the acoustic waves are ultrasound. The receiver has a signal-strength indicator and a servo that turns a directional ultrasonic transducer. There are two signal sources at different frequencies. When the transducer is rotated so the signal from one source is maximum, a bearing is obtained  Displacement Error by comparing the orientation of the transducer with some known standard such as the reading of a magnetic compass. The same is done for the other source. A computer uses triangulation to figure out the precise location of the robot. Radio direction finding (RDF) A radio receiver, equipped with a signal-strength indicator and connected to a rotatable, directional antenna, can be used to determine the direction from which signals are coming. Radio direction finding (RDF) equipment aboard a mobile robot facilitates determining the location of a transmitter. RDF equipment can be used to find the location of a robot with respect to two or more transmitters operating on different frequencies. In an RDF receiver, a loop antenna is generally used. It is shielded against the electric component of radio waves, so it picks up only the magnetic flux. The circumference is less than 0.1 wavelength. The loop is rotated until a dip occurs in the received signal strength. When the dip is found, the axis of the loop lies along a line toward the transmitter. When readings are taken from two or more locations separated by a sufficient distance, the transmitter can be pinpointed by finding the intersection point of the azimuth bearing lines on a map. At frequencies above approximately 300 MHz, a directional transmitting/receiving antenna, such as a Yagi, quad, dish, or helical type, gives better results than a small loop. When such an antenna is employed for RDF, the azimuth bearing is indicated by a signal peak rather than by a dip. See also DIRECTION RESOLUTION and TRIANGULATION. DIRECTION RESOLUTION Direction resolution refers to the ability of a robot to separate two objects that appear, from the robot’s point of view, to lie in almost the same direction. Direction resolution on the Earth’s surface is also called azimuth resolution. Quantitatively, it is specified in degrees, minutes, or seconds of arc. Two objects might be so nearly in the same direction that a robot “sees” them as being one and the same object, but if they are at different radial distances, the robot can tell them apart by distance measurement. See also DIRECTION FINDING, DISTANCE MEASUREMENT, RADAR, and SONAR. DISPLACEMENT ERROR Displacement error refers to an imprecision in robot position that takes place over time. Displacement error can be measured in absolute terms, such as linear units or degrees of arc. It can also be measured in terms of a percentage of the total displacement or rotation.  Displacement Transducer As an example, suppose a mobile robot is programmed to proceed at a speed of 1.500 meters per second (m/s) at an azimuth bearing of 90.00° (due east) on a level surface. After 10 s, this robot can be expected to be 15.00 m due east of its starting position. If the robot encounters an upward incline, the displacement might be less than 15.00 m; if the robot encounters a downslope, the displacement might be more. If the surface banks to the left or the right, the direction of motion can be expected to change, causing the robot to end up to the north or south of its position had it traveled on a level surface. In the ideal scenario, terrain irregularity would not affect the speed or direction of the machine; the displacement error would therefore be zero. Displacement errors can result from the accumulation of kinematic error over time. Compare KINEMATIC ERROR. DISPLACEMENT TRANSDUCER A displacement transducer is a device that measures a distance or angle traversed, or the distance or angle separating two points. Some displacement transducers convert an electrical current or signal into movement over a certain distance or angle. A transducer that measures distance in a straight line is a linear displacement transducer. If it measures an angle, it is an angular displacement transducer. Suppose you want a robot arm to rotate 28° in the horizontal plane— no more and no less. You give a command to the robot controller such as “BR = 28” (base rotation = 28°). The controller sends a signal to the robot arm, so that it rotates clockwise. An angular displacement transducer keeps track of the angle of rotation, sending a signal back to the computer. This signal increases in linear proportion to the angle that the arm has turned. By issuing the command “BR = 28,” you tell the controller two things: 1. Start the base of the arm rotating. 2. Stop the rotation when the arm has turned through 28°. The second component of the command sets a threshold level for the return signal. As the signal from the displacement transducer increases, it reaches this threshold at 28° of rotation. The controller is programmed to stop the arm at this time. There are other ways to get a robot arm to move, besides using displacement transducers. The above is just one example of how such a transducer might be used in a robotic system. See also ROBOT ARM and TEACH BOX. DISTANCE MEASUREMENT Distance measurement, also called ranging, is a scheme that an autonomous robot can use to navigate in its work environment. It also allows a central  Distance Resolution computer to keep track of the whereabouts of insect robots. There are several ways for an autonomous robot to measure the distance between itself and some object. Sonar uses sound or ultrasound, bouncing the waves off of things around the robot and measuring the time for the waves to return. If the robot senses that an echo delay is extremely short, it knows that it is getting too close to something. Acoustic waves propagate at a speed of roughly 335 m/s in dry air at sea level. Radar works like sonar, but uses microwave radio signals rather than sound waves. Light beams can also be used, particularly lasers, in which case the scheme is called ladar. But radio and light beams travel at such high speed (300 million m/s in free space) that it is difficult to measure delay times for nearby objects. Also, some objects reflect light waves poorly, making it difficult to obtain echoes strong enough to allow distance measurement. Stadimetry infers the distance to an object of known height, width, or diameter by measuring the angle the object subtends in the vision system’s field of view. Beacons of various kinds can be used for distance measurement. These devices can use sound, radio waves or light waves. See also AUTONOMOUS ROBOT, BEACON, DISTANCE RESOLUTION, INSECT ROBOT, LADAR, RADAR, SONAR, STADIMETRY, and TIME-OF-FLIGHT DISTANCE MEASUREMENT. DISTANCE RESOLUTION Distance resolution is the precision of a robotic distance measurement system. Qualitatively, it is the ability of the system to differentiate between two objects that are almost, but not quite, the same distance away from the robot. Quantitatively, it can be measured in meters, centimeters, millimeters, or even smaller units. When two objects are very close to each other, a distance-measuring system sees them as a single object. As the objects get farther apart, they become distinguishable. The minimum radial separation of objects, for a ranging system to tell them apart, is the distance resolution. With some distance measuring systems, nearby sets of objects can be resolved better than sets of objects far away. Suppose two objects are separated radially by 1 m. If their mean (average) distance is 10 m, their separation is 1/10 (10 percent) of the mean distance. If their mean distance is 1000 m, their separation is 1/1000 (0.1 percent) of the mean distance. If the distance resolution is 1 percent of the mean distance, then the system can tell the nearer pair of objects apart, but not the more distant pair. Distance resolution depends on the type of ranging system used. The most sensitive methods compare the phases of the wavefronts emitted by laser beams. These waves either arrive from, or are reflected  Distinctive Place by, beacons located at strategic points in the work environment. A highend system of this kind can resolve distances down to a small fraction of a millimeter. See also BEACON, DISTANCE MEASUREMENT, RADAR, and SONAR. DISTINCTIVE PLACE A distinctive place is a point in a mobile robot’s work environment that has special significance, or that can be used as a point of reference for navigational purposes. Such points are determined on the basis of features in specific regions, called neighborhoods, in the work environment. Suppose a mobile robot is designed to function on a single level of an office building. The work environment is the entire floor (the set of all points) over which the machine can move. Each room can be considered a neighborhood. Distinctive places might be defined as the centers of the doorways between adjacent rooms, or between each room and the hallway. Distinctive places might also include the physical (geographic) center of the floor in each room, or the point on the floor that lies at the greatest distance, in any given room, from fixed obstructions. Beacons can also serve as distinctive places. See also BEACON, COMPUTER MAP, and RELATIONAL GRAPH. DISTRIBUTED CONTROL In a system containing more than one robot, distributed control refers to unit independence. In a robotic system that employs distributed control, also known as decentralized control, each robot in the fleet is capable, to some extent, of making its own decisions and operating without instructions from other robots or from a central controller. If there is a central controller, its function is limited. This type of robotic system is analogous to a peer-to-peer computer network. In a robotic system that employs uniformly distributed control, there is no main controller; each robot is fully autonomous, containing its own controller. Each unit is equal to all the others in significance. In some systems, there is a main controller that oversees some of the operations of each unit in the fleet. This is known as partially distributed control. Another example of partially distributed control is a system in which each robot receives a set of instructions from a central controller, stores those instructions, and then carries them out independently of the central controller. In some robotic systems, the individual units are completely and continuously dependent on the central controller, and cannot function if the communication link is severed. Such a system is said to employ fully centralized control. Compare CENTRALIZED CONTROL. See also AUTONOMOUS ROBOT and INSECT ROBOT.  Domain of Function DOMAIN OF FUNCTION The domain of a mathematical function is the set of independent-variable values for which the function is defined. Every x in the domain of a function f is mapped by f onto a definite, single value y. Any x not in the domain is not mapped onto anything by the function f. Suppose you are given the function f (x) = +x 1/2 (that is, the positive square root of x). The graph of this function is shown in the illustration. The function is not defined for negative values of x, and is also not defined, as shown in this particular example, for x = 0. The function f (x) has y 6 4 f ( x) = + x 2 x -6 -4 -2 -2 Domain -4 2 4 6 -6 Domain of function values y if, and only if, x 0. Therefore the domain of f is the set of positive real numbers. Computers work extensively with functions, both analog and digital. Functions are important in robotic navigation, location, and measurement systems. See also FUNCTION and RANGE OF FUNCTION.  Downlink DOWNLINK See UPLINK/DOWNLINK. DROP DELIVERY Drop delivery is a simple method that a robotic end effector can use to place an object into position. The object is picked up by a gripper and then moved until it is directly over a slot, hole, conveyor belt, chute, or other receptacle designed for it. Then the gripper lets go of the object, and it falls into place. Drop delivery requires precision in the movement of the robotic arm and end effector. In addition, when the gripper lets go of the object, it must not impart significant lateral force or torque to the object. Otherwise the object might move out of alignment or topple. If a conveyor belt is used, some means must be employed to ensure that the movement of the belt does not cause the object to slip, tip over, or fall off the belt after it lands. DROPOFF See MAGNITUDE PROFILE. DUTY CYCLE The duty cycle is the proportion of time during which a circuit, machine, or component is operated. Suppose a motor is run for 1 min, then is shut off for 2 min, then is run for 1 min again, and so on. The motor therefore runs for 1 out of every 3 min, or one-third of the time. Its duty cycle is therefore 1⁄3, or 33 percent. If a device is observed for a length of time to, and during this time it runs for a total time t (in the same units as to), then the duty cycle expressed as a percentage, d%, is given by the following formula: d% = 100t to When determining the duty cycle, it is important that the observation time be long enough. In the case of the motor described above, any value of less than 3 min is too short to get a complete sample of the data. Ideally, the observation time should be at least twice the time required for a complete cycle of activity. If the cycle of activity varies somewhat (a common situation), then the observation time must be much greater than the time required for a single cycle. The more a circuit, machine, or component is used, the sooner it will wear out, if all other factors are held constant. In general, the higher the  Dynamic Transducer duty cycle, the shorter is the useful life. This effect is most pronounced when a device is worked near its limits. Also, the rating of a device often depends on the duty cycle at which it is expected to be used. Suppose the motor described above is rated at a torque of 10 newtonmeters (10 N m) for a duty cycle of 100 percent. If the motor is called upon to provide a constant torque of 9.9 N m, then it will be taxed to its utmost. If it must constantly turn a load of 12 N m, it should come as no surprise if it fails prematurely. For a duty cycle of 33 percent, the motor might be rated at 15 N m, as long as any single working period does not exceed 2 min. If it only needs to turn 0.5 N m, the motor can not only run continuously, it will probably last longer than its expected life. Devices such as robot motors can be protected from overwork (either momentary or long-term) by means of back pressure sensors. See BACK PRESSURE SENSOR. DYNAMIC STABILITY Dynamic stability is a measure of the ability of a robot to maintain its balance while in motion. A robot with two or three legs, or that rolls on two wheels, can have excellent stability while it is moving, but when it comes to rest, it is unstable. A two-legged robot can be pushed over easily when it is standing still. This is one of the major drawbacks of biped robots. It is difficult and costly to engineer a good sense of balance, of the sort you take for granted, into a two-legged or two-wheeled machine, although it has been done. Robots with four or six legs have good dynamic stability, but they are usually slower in their movements compared with machines that have fewer legs. See also BIPED ROBOT, INSECT ROBOT, and STATIC STABILITY. DYNAMIC TRANSDUCER A dynamic transducer is a coil-and-magnet device that converts mechanical motion into electricity or vice versa. The most common examples are the dynamic microphone and the dynamic loudspeaker. Dynamic transducers can be used as sensors in a variety of robotic applications. The illustration is a functional diagram of a dynamic transducer suitable for converting sound waves into electric currents or vice versa. A diaphragm is attached to a permanent magnet. The magnet is surrounded by a coil of wire. Acoustic vibrations cause the diaphragm to move back and forth; this moves the magnet, which causes fluctuations in the magnetic field within the coil. The result is alternating-current (AC) output from the coil, having the same waveform as the sound waves that strike the diaphragm.  Dynamic Transducer Diaphragm Coil Magnet Acoustic waves Signal Dynamic transducer If an audio signal is applied to the coil of wire, it creates a magnetic field that produces forces on the permanent magnet. This causes the magnet to move, pushing the diaphragm back and forth. This displaces the air near the diaphragm, producing acoustic waves that follow the waveform of the signal. Dynamic transducers are commonly used in robotic speech recognition and speech synthesis systems. Compare ELECTROSTATIC TRANSDUCER and PIEZOELECTRIC TRANSDUCER. See also SPEECH RECOGNITION and SPEECH SYNTHESIS.  E EDGE DETECTION Edge detection is to the ability of a robotic vision system to locate boundaries. It also refers to a robot’s knowledge of what to do with respect to those boundaries. A robot car, for example, uses edge detection to see the edges of a road, and uses the data to keep itself on the road. However, it also needs to stay a certain distance from the right-hand edge of the pavement, so that it does not cross over into the lane of oncoming traffic. It must stay off the road Sky Grass Road shoulder Blacktop Center line Vision frame Edge detection  Educational Robot shoulder. Thus, it must tell the difference between pavement and other surfaces, such as gravel, grass, sand, and snow. The robot car can use beacons for this purpose, but this requires the installation of the guidance system beforehand, limiting the robot car to roads that are equipped with such navigation aids. A personal robot equipped with edge detection can see certain contours in its work environment. This keeps the machine from running into walls, or closed doors, or windows, or from falling down stairwells. Compare EMBEDDED PATH. See also VISION SYSTEM. EDUCATIONAL ROBOT The term educational robot applies to any robot that causes its user(s) to learn something. Especially, this term applies to robots available for consumer use. Robots of this kind have become popular among children, particularly in Japan, but increasingly in the United States and other Western nations. These machines are toys in the sense that children have fun using them, but they often are excellent teachers as well. Children learn best when they are having fun at the same time. An instructional robot is an educational robot intended to function only, or primarily, as a teacher. Robots of this kind can be purchased for use in the home, but more often they are found in schools, especially at the junior-high and senior-high levels (grades 7 through 12). Robots are intimidating to some students. But once a child or young adult gets used to working or playing with machines, robots can become companions, especially if there is some measure of artificial intelligence (AI). See also PERSONAL ROBOT. ELASTOMER An elastomer is a flexible substance resembling rubber or plastic. In robotic tactile sensing, elastomers can be used to detect the presence or absence of mechanical pressure. The illustration shows how an elastomer can be used to detect, and locate, a pressure point. The elastomer conducts electricity fairly well, but not perfectly. It has a foamlike consistency, so it can be compressed. An array of electrodes is connected to the top of the elastomer pad; an identical array is connected to the bottom of the pad. These electrodes run to the robot controller. When pressure appears at some point in the elastomer pad, the material compresses, and this lowers its electrical resistance in a small region. This is detected as an increase in the current between electrodes in the top pad and the bottom pad, but only within the region where the elastomer is  Electric Eye Electrode array + Battery – Conductive foam Electrode array Current sensors ADC Microcomputer Elastomer being compressed. The data is sent to an analog-to-digital converter (ADC) and then to a microcomputer, which determines where the pressure is taking place, and how intense it is. See also TACTILE SENSING. ELECTRIC EYE An electric eye optically senses an object and then actuates a device. For example, it might be set up to detect anything passing through a doorway. This can count the number of people entering or leaving a building. Another example is the counting of items on a fast-moving assembly line; each item breaks the light beam once, and a circuit counts the number of interruptions. Usually, an electric eye has a light source and a photocell; these are connected to an actuating circuit as shown in the block diagram. When something interrupts the light beam, the voltage or current from the photocell changes dramatically. It is easy for electronic circuits to detect  Electrochemical Power Laser diode Photovoltaic cell Energy beam Lens Lens Power supply Actuating circuit Controlled device Electric eye this voltage or current change. Using amplifiers, even the smallest change can be used to control large machines. Electric eyes do not always operate with visible light. Infrared (IR), with a wavelength somewhat longer than visible red, is commonly used in optical sensing devices. This is ideal for use in burglar alarms, because an intruder cannot see the beam, and therefore cannot avoid it. ELECTROCHEMICAL POWER An electrochemical cell is a unit source of direct-current (DC) power. When two or more such cells are connected in series to increase the voltage, the result is a battery. Electrochemical cells and batteries are extensively used in mobile robots. Lead–acid cell Figure 1 shows an example of a lead–acid cell. An electrode of lead and an electrode of lead dioxide, immersed in a sulfuric acid solution, exhibit a potential difference. This voltage can drive a current through a load. The maximum available current depends on the volume and mass of the cell. If this cell is connected to a load for a long time, the current will gradually decrease, and the electrodes will become coated. The nature of the acid will change. All the potential energy in the acid will be converted into DC electrical energy, and ultimately into heat, visible light, radio waves, sound, or mechanical motion.  Electrochemical Power Primary and secondary cells Some cells, once their chemical energy has all been changed to electricity and used up, must be discarded. These are primary cells. Other kinds of cells, like the lead–acid unit described above, can get their chemical energy back again by means of recharging. Such a component is a secondary cell. Primary cells contain a dry electrolyte paste along with metal electrodes. They go by names such as dry cell, zinc–carbon cell, or alkaline cell. These cells are commonly found in supermarkets and other stores. Some secondary cells can also be found at the consumer level. Nickel–cadmium (Ni–Cd or NICAD) cells are one common type. These cost more than ordinary dry cells, and a charging unit also costs a few dollars. However, these rechargeable cells can be used hundreds of times, and can pay for themselves and the charger several times over. An automotive battery is made from secondary cells connected in series. These cells recharge from an alternator or from an outside charging unit. This type of battery has cells like the one shown in Fig. 1. It is dangerous to short-circuit the terminals of such a battery because the acid can boil out. In fact, it is unwise to short-circuit any cell or battery, because it might explode or cause a fire. Voltage Lead electrode Lead-dioxide electrode Acid solution Electrochemical power—Fig. 1 Storage capacity Common units of electrical energy are the watt hour (Wh) and the kilowatt hour (kWh). Any cell or battery has a certain amount of electrical energy that can be specified in watt hours or kilowatt hours. Often it is given in terms of the mathematical integral of deliverable current with respect to  Electrochemical Power time, in units of ampere hours (Ah). The energy capacity in watt hours is the ampere-hour capacity multiplied by the battery voltage. A battery with a rating of 20 Ah can provide 20 A for 1 h, or 1 A for 20 h, or 100 mA (100 milliamperes) for 200 h. The limitations are shelf life at one extreme, and maximum deliverable current at the other. Shelf life is the length of time the battery will remain usable if it is never connected to a load; this is measured in months or years. The maximum deliverable current is the highest current a battery can drive through a load without the voltage dropping significantly because of the battery’s own internal resistance. Small cells have storage capacity of a few milliampere hours (mAh) up to 100 or 200 mAh. Medium-sized cells might supply 500 mAh to 1000 mAh (1 Ah). Large automotive lead–acid batteries can provide upwards of 100 Ah. Discharge curve When an ideal cell or ideal battery is used, it delivers a constant current for a while, and then the current starts to decrease. Some types of cells and batteries approach this ideal behavior, exhibiting a flat discharge curve (Fig. 2). Others have current that decreases gradually from the beginning of use; this is a declining discharge curve (Fig. 3). Relative current Useful life Low battery Time Electrochemical power—Fig. 2  Electrochemical Power Relative current Fresh battery Low battery Time Electrochemical power—Fig. 3 When the current that a battery can provide has decreased to about half of its initial value, the cell or battery is said to be “weak” or “low.” At this time, it should be replaced. A battery should not be allowed to run down until the current drops to nearly zero. Common cells and batteries The cells sold in stores, and used in convenience items such as flashlights and transistor radios, are usually of the zinc–carbon or alkaline variety. These provide 1.5 volts (V) and are available in sizes AAA (very small), AA (small), C (medium), and D (large). Batteries made from these cells are usually rated at 6 V or 9 V. Zinc–carbon cells have a fairly long shelf life. The zinc forms the outer case and is the negative electrode. A carbon rod serves as the positive electrode. The electrolyte is a paste of manganese dioxide and carbon. Zinc–carbon cells are inexpensive and are usable at moderate temperatures, and in applications where the current drain is moderate to high. They do not work well in extremely cold environments. Alkaline cells have granular zinc for the negative electrode, potassium hydroxide as the electrolyte, and a polarizer as the positive electrode. An alkaline cell can work at lower temperatures than a zinc–carbon cell. It  Electrochemical Power also lasts longer in most electronic devices, and is therefore preferred for use in transistor radios, calculators, and portable cassette players. Its shelf life is much longer than that of a zinc–carbon cell. Transistor batteries are small, 9-V, box-shaped batteries with clip-on connectors on top. They consist of six tiny zinc–carbon or alkaline cells in series. These batteries are used in low-current electronic devices, such as portable earphone radios, radio garage-door openers, television and stereo remote-control boxes, and electronic calculators. Lantern batteries are rather massive, and can deliver a fair amount of current. One type has spring contacts on the top. The other type has thumbscrew terminals. Besides keeping an incandescent bulb lit for a while, these batteries, usually rated at 6 V and consisting of four zinc–carbon or alkaline cells, can provide enough energy to operate a low-power communications radio or a small mobile robot. Silver-oxide cells are usually made into a buttonlike shape, and can fit inside a wristwatch. They come in various sizes and thicknesses, all with similar appearance. They supply 1.5 V and offer excellent energy storage for the weight. They have a flat discharge curve. Silver-oxide cells can be stacked to make batteries about the size of an AA cylindrical cell. Mercury cells, also called mercuric oxide cells, have advantages similar to silver-oxide cells. They are manufactured in the same general form. The main difference, often not of significance, is a somewhat lower voltage per cell: 1.35 V. There has been a decrease in the popularity of mercury cells and batteries in recent years, because mercury is toxic and is not easily disposed of. Lithium cells supply 1.5 to 3.5 V, depending on the chemistry used. These cells, like their silver-oxide cousins, can be stacked to make batteries. Lithium cells and batteries have superior shelf life, and they can last for years in very-low-current applications. They provide excellent energy capacity per unit volume. Lead–acid cells and batteries have a solution or paste of sulfuric acid, along with a lead electrode (negative) and a lead-dioxide electrode (positive). Paste-type lead–acid batteries can be used in consumer devices that require moderate current, such as laptop computers, portable VCRs, and personal robots. They are also used in uninterruptible power supplies for personal computers. Nickel-based cells and batteries NICAD cells come in several forms. Cylindrical cells look like dry cells. Button cells are used in cameras, watches, memory backup applications, and other places where miniaturization is important. Flooded cells are used in heavy-duty applications, and can have a storage capacity of as  Electromechanical Transducer much as 1000 Ah. Spacecraft cells are made in packages that can withstand extraterrestrial temperatures and pressures. NICAD batteries are available in packs of cells that can be plugged into equipment to form part of the case for a device. An example is the battery pack for a hand-held radio transceiver. NICAD cells and batteries should never be left connected to a load after the current drops to zero. This can cause the polarity of a cell, or of one or more cells in a battery, to reverse. Once this happens, the cell or battery will no longer be usable. When a NICAD nears full discharge, it should be recharged as soon as possible. Nickel–metal-hydride (NiMH) cells and batteries can directly replace NICAD units in most applications. See also POWER SUPPLY and SOLAR POWER. ELECTROMAGNETIC SHIELDING Electromagnetic shielding is a means of preventing computers and other sensitive equipment from being affected by stray electromagnetic (EM) fields. Computers also generate EM energy of their own, and this can cause interference to other devices, especially radio receivers, unless shielding is used. The simplest way to provide EM shielding for a circuit is to surround it with metal, usually copper or aluminum, and to connect this metal to electrical ground. Because metals are good conductors, an EM field sets up electric currents in them. These currents oppose the EM field, and if the metal enclosure is grounded, the EM field is in effect shorted out. Interconnecting cables should also be shielded for optimum protection against electromagnetic interference (EMI). This is done by surrounding all the cable conductors with a copper braid. The braid is electrically grounded through the connectors at the ends of the cable. One of the biggest advantages of fiber-optic data transmission is the fact that it does not need EM shielding. Fiber-optic systems are immune to the EM fields produced by radio transmitters and alternating-current (AC) utility wiring. Fiber-optic systems also work without generating external EM fields, so they do not cause EMI to surrounding circuits and devices. ELECTROMECHANICAL TRANSDUCER An electromechanical transducer is a device that converts electrical energy into mechanical energy or vice versa. Electric motors and electric generators are the most common examples. A motor works by means of magnetic forces produced by electric currents; the generator produces electric currents as a result of the motion of an electric conductor in a magnetic field.  Electrostatic Transducer Devices that convert sound into electricity, or vice versa, are another form of electromechanical transducer. Speakers and microphones are universal examples. They usually work by means of dynamic principles, but some work by electrostatic interactions. Galvanometer-type analog meters, also known as D’Arsonval meters, are electromechanical transducers. They convert electric current into displacement. In recent years, digital meters have largely replaced electromechanical meters. Digital devices do not have moving parts to wear out, so they last much longer than electromechanical types. Digital meters are also able to tolerate more physical abuse. Robots use electromechanical transducers in many ways. Examples include the selsyn, the stepper motor, and the servomechanism. See also SELSYN, STEPPER MOTOR, and SERVOMECHANISM. ELECTROSTATIC TRANSDUCER An electrostatic transducer is a device that changes mechanical energy into electrical energy or vice versa by taking advantage of electrostatic forces. The most common types involve conversion between sound waves and audio-frequency electric currents. The illustration is a functional diagram of an electrostatic transducer. It can function either as a microphone (sound-to-current transducer) or a speaker (current-to-sound transducer). In the “microphone mode,” incoming sound waves cause vibration of the flexible plate. This produces rapid (although small) changes in the Rigid plate Blocking capacitor Signal Acoustic waves Flexible plate DC power source Electrostatic transducer  Empirical Design spacing, and therefore the capacitance, between the two plates. A directcurrent (DC) voltage is applied to the plates, as shown. As the capacitance changes between the plates, the electric field between them fluctuates. This produces variations in the current through the primary winding of the transformer. Audio signals appear across the secondary winding. In “speaker mode,” currents in the transformer produce changes in the voltage between the plates. This change results in electrostatic force fluctuations, pulling and pushing the flexible plate in and out. The motion of the flexible plate produces sound waves. Electrostatic transducers can be used in most applications where other types of transducers are employed. This includes speech recognition and speech synthesis systems. Advantages of electrostatic transducers include light weight and excellent sensitivity. They can also work with small electric currents. Compare DYNAMIC TRANSDUCER and PIEZOELECTRIC TRANSDUCER. See also SPEECH RECOGNITION and SPEECH SYNTHESIS. EMBEDDED PATH An embedded path is a means of guiding a robot along a specific route. The automated guided vehicle (AGV) employs this scheme. One common type of embedded path is a buried, current-carrying wire. The current in the wire produces a magnetic field that the robot can follow. This method of guidance has been suggested as a way to keep a car on a highway, even if the driver does not pay attention. The wire needs a constant supply of electricity for this guidance method to work. If the current is interrupted for any reason, the robot will lose its way. Alternatives to wires, such as colored paints or tapes, do not need a supply of power, and this gives them an advantage. Tape is easy to remove and put somewhere else; this is difficult to do with paint, and practically impossible with wires embedded in concrete. Compare EDGE DETECTION. See also AUTOMATED GUIDED VEHICLE. EMPIRICAL DESIGN Empirical design is an engineering technique in which experience and intuition are used in addition to theory. The process is largely trial and error. The engineer starts at a logical point, based on theoretical principles, but experimentation is necessary in order to get the device or system to work just right. Robots are ideally suited to empirical design techniques. An engineer cannot draw up plans for a robot, no matter how detailed or painstaking the drawing-board process might be, and expect the real machine to work perfectly on the first trial. A prototype is built and tested, noting the flaws. The engineer goes back to the drawing board and revises the design. Sometimes it is necessary to start all over from scratch; more often, small  End Effector changes are made. The machine is tested again, and the problems noted. Another drawing-board round follows. This process is repeated until the machine works the way the engineer (or the customer) wants. END EFFECTOR An end effector is a device or tool connected to the end of a robot arm. The nature of the end effects depends on the intended task. If a robot is designed to set the table for supper, “hands,” more often called robot grippers, can be attached to the ends of the robot arms. In an assembly-line robot designed to insert screws into cabinets, a rotatingshaft device and screwdriver head can be attached at the end of the arm. Such a rotating shaft might also be fitted with a bit for drilling holes, or an emery disk for sanding wood. A given type of robot arm can usually accommodate only certain kinds of end effectors. One cannot take a table-setting robot, simply replace one of its grippers with a screwdriver, and then expect it to tighten the screws on the hinges of kitchen cabinets. Such a task change requires a change in the programming of the robot controller, so it operates in “handyrobot mode” rather than in “waitrobot mode.” One must also change the hardware in the robot arm to operate a rotating end effector rather than a gripper. See also ROBOT ARM and ROBOT GRIPPER. ENTITIZATION Entitization, also called objectization, is an expression of the ease with which a robot can differentiate among objects in its work environment. It is an indication of the effectiveness of object recognition, and can be defined in qualitative or quantitative terms. Qualitative expressions of entitization are adjectives (such as “good,” “fair,” or “poor”). Quantitative entitization is determined on the basis of the proportion of correct identifications in a large number of tests in a practical scenario. For example, if a robot correctly identifies an object 997 out of 1000 times, its entitization is 99.7 percent accurate. See also OBJECT RECOGNITION. ENVELOPE See WORK ENVELOPE. EPIPOLAR NAVIGATION Epipolar navigation is a means by which a machine can locate objects in three-dimensional (3-D) space. It can also navigate, and figure out its own position and path. Epipolar navigation works by evaluating the way  Epipolar Navigation an image appears to change as viewed from a moving point of view. The human eye/brain system does this to a limited degree with little thought or conscious effort. Robot vision systems can do it with extreme precision. To illustrate epipolar navigation, imagine a robotic aircraft (drone) flying over the ocean. The only land beneath the drone is a small island (see the illustration). The robot controller has, on its hard drive, an excellent map that shows the location, size, and exact shape of this island. For instrumentation, the drone has only a computer, a good video camera, and sophisticated programming. The drone can navigate its way by Sighting A Sighting B Sighting C Island Ocean Sighting A Sighting C Sighting B Epipolar navigation  Error Accumulation observing the island and scrutinizing the shape and angular size of the island’s image. As the drone flies along, the island seems to move underneath it. A camera is fixed on the island. The controller sees an image that constantly changes shape and angular size. The controller is programmed with the true size, shape, orientation, and geographic location of the island. The controller compares the shape/size of the image it sees, from the vantage point of the aircraft, with the actual shape/size of the island, that it “knows” from the map data. From this alone, it can precisely determine the drone’s: • • • • • Altitude Speed of travel relative to the surface Direction of travel relative to the surface Geographic latitude Geographic longitude The key is that there exists a one-to-one correspondence between all points within sight of the island, and the angular size and shape of the island’s image. The correspondence is far too complex for a human being to memorize exactly; but for a computer, matching the image it sees with a particular point in space is easy. Epipolar navigation can, in theory, work on any scale, and at any speed—even relativistic speeds at which time dilation occurs. It is a method by which robots can find their way without triangulation, direction finding, beacons, sonar, or radar. It is, however, necessary that the robot have a detailed, precise, and accurate computer map of its environment. See also COMPUTER MAP, LOG POLAR NAVIGATION, and VISION SYSTEM. ERROR ACCUMULATION When measurements are made in succession, the maximum possible error adds up. This is called error accumulation. Analog error accumulation can be illustrated by a measurement example. Suppose you want to measure a long piece of string (about 100 m, say), using a meter stick marked off in millimeters. You must place the stick along the string over and over again, about 100 times. If your error is up to ±2 mm with each measurement, then after 100 repetitions, the possible error is up to ±200 mm. Digital error accumulation occurs as bits are misread in a communications circuit, incorrectly written on disk, or incorrectly stored in memory. A machine might see a logic low when it should see high, or vice versa. Suppose that, for a particular digital file, an average of three errors are introduced each time the file is transferred from one node to another in  Error-Sensing Circuit a communication circuit. If the signal passes through n nodes, there will be an average of 3n (3 + 3 + 3 + + 3, n times) errors. In robotic systems, kinematic errors, or errors in movement, can accumulate over time, resulting in eventual positioning or displacement errors. See also KINEMATIC ERROR. ERROR CORRECTION Error correction is a form of computer programming in which certain types of mistakes are corrected automatically. An example is a program that maintains a large dictionary of English words. The operator of a computer connected to a speech-synthesizing robot might misspell words or make typographical errors. Running the error-correction program will cause the computer to single out all peculiar-looking words, bringing them to the attention of the operator. The operator can then decide whether the word is correct. With modern computers, huge vocabularies are easily stored. When robots must keep track of variables such as position and speed, error correction can be used when an instrument is known to be imprecise, or when values depart from the reasonable range. A computer can keep track of error accumulation, checking periodically to be sure that discrepancies are not adding up beyond a certain maximum. Error correction is important in robotic systems subject to gravity loading. In order to ensure that the end effector in a robot arm does not stray from its intended position because of the force of gravity on the assembly itself, position sensing devices can be used, and a feedback system employed to counter-move the robot arm until the error signal from the sensor is zero. In robotic navigation systems, error correction refers to the set of processes that keep the device on its intended course. In a servomechanism, error correction is done by means of feedback. See also ERROR ACCUMULATION, ERROR SENSING CIRCUIT, ERROR SIGNAL, POSITION SENSING, and SERVOMECHANISM. ERROR-SENSING CIRCUIT An error-sensing circuit produces a signal when two inputs are different, or when a variable deviates from a chosen value. If the two inputs are the same, or if the variable is at the chosen value, the output is zero. This type of circuit is also sometimes called a comparator. Suppose you want a robot to home in on some object. The object has a radio transmitter that sends out a beacon signal. The robot has radio direction finding (RDF) equipment built-in. When the robot is heading in the right direction, the beacon is in the RDF null, and the received signal strength is zero, as shown in the accompanying polar-coordinate plot. If  Error Signal DF response curve North Robot To beacon East South Error-sensing circuit the robot turns off course, the beacon is no longer in the null, and a signal is picked up by the RDF receiver. This signal goes to the robot controller, which steers the robot to the left and right until the beacon signal once again falls into the null. See also DIRECTION FINDING and SERVOMECHANISM. ERROR SIGNAL An error signal is a voltage generated by an error-sensing circuit. This signal occurs whenever the output of the device differs from a reference value. Error signals can be used in purely electronic systems, and also in electromechanical systems. In the RDF device described under ERROR-SENSING CIRCUIT, the output might look like the polar-coordinate graph shown in the illustration. If the robot is pointed on course, the error signal is zero. If it is off course, either to the left or the right, a positive error signal voltage is generated, as shown in the accompanying rectangular-coordinate graph. The voltage depends on how far off course the robot is headed. In general, as the heading error increases, so does the error-signal strength.  Exoskeleton Signal strength Too far left Error signal On course Too far right A direction-finding circuit is designed to seek out, and maintain, a heading such that the error signal is always zero. To do this, the error signal is used by the robot controller to change the heading. This is the same principle by which a hidden radio transmitter is found. See also BEACON, DIRECTION FINDING, ERROR CORRECTION, and SERVOMECHANISM. EXOSKELETON An exoskeleton is a robot arm that uses articulated geometry to mimic the motions of a human arm, and whose motions are controlled directly by movements of the arm of a human operator. Such devices can be used when working with hazardous materials. They are also useful as prostheses (artificial limbs). See ARTICULATED GEOMETRY and PROSTHESIS. The term exoskeleton also refers to a specialized robot that is like a suit of armor a human can wear, and which can amplify movement displacement and/or force, resulting in physical strength far beyond that of an ordinary man or woman. A woman might, for example, lift a car over her head; the steel frame of the exoskeleton would bear the weight and pressure. A man might throw a baseball a kilometer. The armor could protect against blows, fire, and perhaps even bullets. Full exoskeletons have, to date, been implemented mainly in science-fiction stories.  Expert System A full exoskeleton differs from a telepresence system. In telepresence, the human operator is not at the same location as the robot. But when a human wears an exoskeleton, he or she is on site with the machine. This is both an asset and a liability: it allows for greater control and better sensing of the work environment, but in some instances it can place the human operator in physical peril. Compare TELEPRESENCE. EXPERT SYSTEM An expert system is a scheme of computer reasoning, also known as a rule-based system. Expert systems are used in the control of smart robots. They can also be employed in stand-alone computers. The drawing is a block diagram of a typical expert system. The heart of the device is a set of facts and rules. In the case of a robotic system, the facts consist of data about the robot’s environment, such as a factory, an office, or a kitchen. The rules are statements of the logical form “If X, then Y,” similar to statements in high-level programming languages. An inference engine decides which logical rules should be applied in various situations. Then it instructs the robot(s) to carry out certain tasks. However, the operation of the system can only be as sophisticated as the data supplied by human programmers. Robot Rules Robot Inference engine Robot Human operator Facts Robot Expert system  Eye-In-Hand System Expert systems can be used in computers to help people do research, make decisions, and generate forecasts. A good example is a program that assists a physician in making a diagnosis. The computer asks questions, and arrives at a conclusion based on the answers given by the patient and doctor. One of the biggest advantages of expert systems is the fact that reprogramming is easy. As the environment changes, the robot can be taught new rules, and supplied with new facts. EXTENSIBILITY Extensibility, also called expandability, refers to the ease with which a robotic system can be modified to perform a greater number, or a greater variety, of tasks than those allowed for in its original design. The extensibility of a robotic system depends on various factors, including the nature of the hardware, the controller memory, the controller data storage space, and the controller processing speed. Extensibility is enhanced by the use of modular construction and standardized parts. EXTRAPOLATION When data are available within a certain range, an estimate of values outside that range can be made by a technique called extrapolation. This can be educated guessing, but it can also be done using a computer. The more sophisticated the computer software, the more accurately it can extrapolate. An example of extrapolation is the forecast path of a hurricane as it approaches a coastline. Knowing its path up to the present moment, a range of possible future paths is developed by the computer. Factors that can be programmed into the computer to help it make an accurate extrapolation include: • Paths of hurricanes in past years that approached in a similar way • Steering currents in the upper atmosphere • Weather conditions in the general path of the storm The farther out (into the future) an extrapolation is made, the less accurate the results. While a weather computer might do a good job of predicting a hurricane path 24 h in advance, no machine yet devised can tell exactly where the storm will be in a week. Compare INTERPOLATION. EYE-IN-HAND SYSTEM For a robot gripper to find its way, a camera can be placed in the gripper mechanism. The camera must be equipped for work at close range, from about 1 m down to a few millimeters. The positioning error must be as small as possible, preferably less than 0.5 mm. To be sure that the camera  Eye-In-Hand System gets a good image, a lamp is included in the gripper along with the camera (see the drawing). The so-called eye-in-hand system can be used to measure precisely how close the gripper is to whatever object it is seeking. It can also make positive identification of the object, so that the gripper does not go after the wrong thing. The eye-in-hand system uses a servomechanism. The robot is equipped with, or has access to, a controller that processes the data from the camera and sends instructions back to the gripper. See also FINE MOTION PLANNING and ROBOT GRIPPER. Robot arm Lamp Lens Lamp Camera Grippers Eye-in-hand system  F FALSE NEGATIVE OR POSITIVE Sensors do not always react as intended to stimuli, or percepts, in the environment. This can occur for a variety of reasons, and is known as a false negative. Conversely, sensors occasionally produce output when no legitimate percept is present; this is a false positive. Consider an infrared (IR) sensor. Suppose it is most sensitive at a wavelength of 1350 nm (nanometers). False negatives are least likely to occur for percepts at that wavelength. As the wavelength departs from 1350 nm, the sensitivity decreases, and the radiation must be more intense to cause the sensor to produce an output signal. The likelihood of false negatives increases as the wavelength becomes longer or shorter than 1350 nm. Outside a certain range of wavelengths, the sensor is relatively insensitive, and false negatives are therefore the rule rather than the exception. Whether the failure to produce output constitutes a false negative depends, however, on the range of wavelengths that are defined as “legitimate” percepts. Suppose the sensor in the foregoing example is part of a proximitydetection device on a mobile robot. A laser on the robot, operating at a wavelength of 1350 nm, reflects from nearby objects in the work environment. The reflections are picked up by the sensor, which is covered by an IR filter that passes radiation easily within the range 1300 to 1400 nm, but blocks most energy outside that range. If the sensor output exceeds a certain level, the robot controller is instructed to change direction to avoid striking a possible obstruction. External sources of IR can cause false positives. This is most likely to occur if the external IR has a wavelength near the peak sensitivity region of the sensor/filter, that is, between 1300 and 1400 nm. However, if the external percept is sufficiently intense, it might cause a false positive even if its wavelength is considerably less than 1300 nm or greater than 1400 nm. Robot controllers can be programmed to ignore false negatives or positives, as long as there is some way to distinguish between them and  False Resilience “legitimate” percepts. In a poorly designed system, however, false negatives or positives can cause erratic operation. FAULT RESILIENCE The term fault resilience can refer to either of two different characteristics of a computerized robotic system. The first type of fault-resilient system can also be called sabotage-proof. Suppose that all the strategic (nuclear) defenses of the United States are placed under the control of a computer. It is imperative that it be impossible for unauthorized people to turn it off. Backup systems are necessary. No matter what anyone tries to do to cause the system to malfunction or become inoperative, the system must be capable of resisting or overcoming such attack. Some engineers doubt that it is possible to build a totally sabotageproof computer. They quote the saying, “Build a more crime-proof system, and you get smarter criminals.” Also, any such system would have to be engineered and built by human beings. At least one of those people could be bribed or blackmailed into divulging information on how to defeat the security provisions. And of course, no one can anticipate all of the things that might go wrong with a system. According to Murphy’s law, which is usually stated tongue-in-cheek but which can often manifest itself as truth, “If something can go wrong, it will.” And the corollary, less often heard but perhaps just as true, is “If something cannot go wrong, it will.” The second type of fault resilience is also known as graceful degradation. Many computers and also computer-controlled robotic systems are designed so that if some parts fail, the system still works, although perhaps at reduced efficiency and speed. See GRACEFUL DEGRADATION. FEEDBACK Feedback is a means by which a closed-loop system regulates itself. Feedback is used extensively in robotics. An example of feedback can be found in a simple thermostat mechanism, connected to a heating/cooling unit. Suppose the thermostat is set for 20 degrees Celsius (20°C). If the temperature rises much above 20°C, a signal is sent to the heating/cooling unit, telling it to cool the air in the room. If the temperature falls much below 20°C, a signal tells the unit to heat the room. This process is illustrated in the block diagram. In a system that uses feedback to stabilize itself, there must be some leeway between the opposing functions. In the case of the thermostatically controlled heating/cooling system, if both thresholds are set for exactly 20°C, the system will constantly and rapidly cycle back and forth between  Fiber-Optic Cable Measure temperature Much above 20°C? Activate heating unit No Yes Activate cooling unit Yes Much below 20°C? No Feedback heating and cooling. A typical range might be 18 to 22°C. The leeway should, however, not be too wide. See also SERVOMECHANISM. FIBER-OPTIC CABLE A fiber-optic cable is a bundle of transparent, solid strands designed to carry modulated light or infrared (IR). This type of cable can carry millions of signals at high bandwidth. Manufacture Optical fibers are made from glass to which impurities have been added to maximize the transparency at certain wavelengths. The impurities also optimize the refractive index of the glass, or the extent to which it slows  Fiber-Optic Cable Cladding Core Y X Step-index fiber Cladding Core Y X Graded fiber Fiber-optic cable down and bends light. An optical fiber has a core surrounded by a tubular cladding, as shown in the illustrations. The cladding has a lower refractive index than the core. In a step-index optical fiber (top drawing), the core has a uniform index of refraction and the cladding has a lower index, also uniform. The transition at the boundary is abrupt. In the graded-index optical fiber (lower drawing), the core has a refractive index that is greatest along the central axis and steadily decreases outward from the center. At the boundary, there is an abrupt drop in the refractive index. Operation In the top illustration, showing a step-index fiber, ray X enters the core parallel to the fiber axis and travels without striking the boundary unless there is a bend in the fiber. If there is a bend, ray X veers off center and behaves like Y. Ray Y strikes the boundary repeatedly. Each time ray Y encounters the boundary, total internal reflection occurs, so ray Y stays within the core.  Field of View (FOV) In the lower drawing, showing a graded-index fiber, ray X enters the core parallel to the fiber axis and travels without striking the boundary unless there is a bend in the fiber. If there is a bend, ray X veers off center and behaves like ray Y. As ray Y moves farther from the center of the core, the index of refraction decreases, bending the ray back toward the center. If ray Y enters at a sharp enough angle, it might strike the boundary, in which case total internal reflection occurs. Therefore, ray Y stays within the core. Bundling Optical fibers can be bundled into cable, in the same way as wires are bundled. The individual fibers are protected from damage by plastic jackets. Common coverings are polyethylene and polyurethane. Steel wires or other strong materials are often used to add strength to the cable. The whole bundle is encased in an outer jacket. This outer covering can be reinforced with wire mesh and/or coated with corrosionresistant compounds. Each fiber in the bundle can carry several rays of visible light and/or infrared (IR), each ray having a different wavelength. Each ray can in turn contain a large number of signals. Because the frequencies of visible light and IR are much higher than the frequencies of radio-frequency (RF) currents, the bandwidth of an optical/IR cable link can be far greater than that of any RF cable link. This allows much higher data speed. FIELD OF VIEW (FOV) The field of view (FOV) of a directional sensor is a quantitative expression of the angular range within which it reacts properly to stimuli, or percepts. The FOV is defined in terms of x and y (major and minor) angles, and applies primarily to unidirectional sensors (that is, devices intended to pick up energy from one direction). These angles can be defined as radial, relative to the axis at which the sensor is most responsive, or diametric (twice the radial value). The horizontal FOV of a sensor takes the shape of a cone in threedimensional (3-D) space, with the apex at the sensor, as shown in the illustration. This cone does not necessarily have the same flare angle in all planes passing through its axis. As “seen” from the point of view of the sensor itself, the cone appears as a circle or ellipse in an image of the work environment. If the FOV cone is circular, then the x and y angles are the same. If the FOV cone is elliptical, then the x and y angles differ. Compare RANGE.  Fine Motion Planning Response cone Sensor y x Field of view (FOV) FINE MOTION PLANNING Fine motion planning refers to the scheme used by a robot to get into exactly the right position. Suppose a personal robot is told to switch on the light in a hallway. The light switch is on the wall. The robot controller has a computer map of the house, and this includes the location of the hallway light switch. The robot proceeds to the general location of the switch, and reaches for the wall. How does it know exactly where to find the switch, and how to position its gripper precisely so that it will move the toggle on the switch? One method is to incorporate robot vision, such as an eye-in-hand system. This allows the robot to recognize the shape of the toggle and guide itself accordingly. Another method involves the use of tactile sensing, so the end effector can “feel” along the wall in a manner similar to the way a human would find and actuate the switch with eyes closed. Yet another scheme might involve a highly precise, scaled-down epipolar navigation scheme. Compare GROSS MOTION PLANNING. See also COMPUTER MAP, EPIPOLAR NAVIGATION, EYE-IN-HAND SYSTEM, TACTILE SENSING, and VISION SYSTEM. FIRE-PROTECTION ROBOT One role for which robots are especially well suited is fire fighting. If all fire fighters were robots, there would be no risk to human life in this occupation. Robots can be built to withstand far higher temperatures than humans can tolerate. Robots do not suffer from smoke inhalation. The main challenge is to program the robots to exercise judgment as keen as that of human beings, in a wide variety of situations.  Flexible Automation One way to operate fire-protection robots is to have human operators at a remote point, and to equip the machines with telepresence. The operator sits at a set of controls, or wears a full-body suit with controls incorporated. When the operator moves a certain way, the robot moves in exactly the same way. Television cameras in the robot transmit images to the operator. The operator can “virtually” go where the robot goes, without any of the attendant risk. One of the primary duties of household personal robots is to ensure the safety of the human occupants. This must include escorting people from the house if it catches fire, and then putting out the fire and/or calling the fire department. It might also involve performing some first-aid tasks. See also TELEPRESENCE. FIRMWARE Firmware is a term referring to computer programs that are permanently installed in a system. Usually this is done in read-only memory (ROM). The firmware in a computer can be altered, but this requires a hardware change. This might mean physically replacing an integrated circuit (IC), but there are devices whose firmware can be erased and then reprogrammed. These are called erasable programmable read-only memory (EPROM) ICs. Special equipment is needed to change the contents of an EPROM. Firmware programming is common in microcomputer-controlled appliances and machinery, such as fixed-sequence robots that perform a given task repeatedly. Compare HARD WIRING. FIXED-SEQUENCE ROBOT A fixed-sequence robot is a robot that performs a single, preprogrammed task or set of tasks, making exactly the same movements each time. There is no exception or variation to the routine. Fixed-sequence robots are ideally suited to assembly-line work. An entertaining example of a fixed-sequence robot is a toy that goes through some routine whenever a button is pressed. These machines are especially popular in Japan. In some cases, such toy robots appear sophisticated. FLEXIBLE AUTOMATION Flexible automation refers to the ability of a robot or system to do various tasks. To change from one task to another, a simple software change, or a change in the commands input to the controller, is all that is necessary. A simple example of flexible automation is a robot arm that can be programmed to insert screws, drill holes, sand, weld, insert rivets, and spray paint on objects in an assembly line.  Flight Telerobotic Servicer As personal robots evolve, they become capable of doing many things on the basis of a single, sophisticated program. This is the ultimate in flexible automation, and can be considered a form of artificial intelligence (AI). The appropriate actions result from verbal commands. This necessitates speech-recognition capabilities, as well as considerable controller memory, speed, and processing power. FLIGHT TELEROBOTIC SERVICER In space missions, it is often necessary to perform repairs and general maintenance in and around the spacecraft. It is not always economical to have astronauts do this work. For this reason, various designs have been considered for a robot called a flight telerobotic servicer (FTS). Camera with lamp Power pack Right arm and gripper Left arm and gripper Leg Base gripper Flight telerobotic servicer  Fluxgate Magnetometer The FTS is a remote-controlled robot. The extent to which it is controlled depends on the design. The simplest FTS machines are programmable from the spacecraft’s main computer. More complex FTS devices make use of telepresence. Because of the risk involved in sending humans into space, scientists have considered the idea of launching FTS-piloted space shuttles to deploy or repair satellites. The FTSs would be controlled through computers on the ground and in the spacecraft. One FTS design has the appearance of a one-legged, headless android, as shown in the illustration. See also TELEPRESENCE. FLOWCHART A flowchart is a diagram that illustrates a logical process or a computer program. It is a block diagram. Boxes indicate conditions, diamonds indicate decision points, and arrows show procedural steps. Flowcharts are used to develop computer software. They are also used in troubleshooting of complex equipment. Flowcharts lend themselves well to robotic applications, because they indicate choices that a robot must make while it accomplishes a task. A flowchart must always represent a complete process. There should be no places where a technician, computer, or robot will be left without some decision being made. There must be no infinite loops, where the process goes in endless circles without accomplishing anything. Examples of flowcharts are shown in the definitions of BRANCHING and FEEDBACK. FLUXGATE MAGNETOMETER A fluxgate magnetometer is a computerized robot guidance system that uses magnetic fields to derive position and orientation data. The device uses coils to sense changes in the geomagnetic field (Earth’s magnetic field), or in an artificially generated reference field. Navigation within a defined area can be carried out by having the robot controller constantly analyze the orientation and intensity of the magnetic flux field generated by strategically placed electromagnets. A computer map of the flux field, showing two electromagnets and a hypothetical robot in the field, is shown in the illustration. In this case, opposite magnetic poles (north and south) face each other, giving the flux field a characteristic bar-magnet shape. For each point in the work environment, the magnetic flux has a unique orientation and intensity. Therefore, there is a one-to-one correspondence between magnetic flux conditions and each point inside the environment. The robot controller is programmed to “know” this  Flying Eyeball Robot Magnet Magnet Magnetic lines of flux Fluxgate magnetometer relation precisely for all points in the environment. This allows the robot to pinpoint its position in three-dimensional (3-D) space, provided a set of reference coordinates is established. See also COMPUTER MAP. FLYING EYEBALL The flying eyeball is a simple form of submarine robot. This robot can resolve detail underwater, and can also move around. It cannot manipulate anything; it has no robot arms or end effectors. Flying eyeballs are used in scientific and military applications. A cable, containing the robot in a special launcher housing, is dropped from a boat. When the launcher gets to the desired depth, it lets out the robot, which is connected to the launcher by a tether, as shown in the illustration. The tether and the drop cable convey data back to the boat. The robot contains a video camera and one or more lamps to illuminate the undersea environment. It also has a set of thrusters, or propellers, that let it move around according to control commands sent through the cable and tether. Human operators on board the boat can watch the images  Food-Service Robot To boat Drop cable Thruster Camera Launch box Lamp Thruster Flying eyeball from the television camera, and guide the robot around as it examines objects on the sea floor. In some cases, the tether can be eliminated, and radio-frequency (RF), infrared (IR), or visible beams can be used to convey data from the robot to the launcher. This allows the robot to have enhanced freedom of movement, without the concern that the tether might get tangled up in something. However, the range of the RF, IR, or link is limited, because water does not propagate these forms of energy for long distances. FOOD-SERVICE ROBOT Robots can be used to prepare and serve food. The major applications are in repetitive chores, such as placing measured portions on plates, assemblyline style, to serve a large number of people. Food-service robots are also used in canning and bottling plants, because these jobs are simple, repetitive, mundane, and easily programmable. As a row of bottles goes by, for example, one robot fills each bottle. Then a machine checks to be sure each bottle is filled to the right level. Rejects are thrown out by another robot. Still another robot places the caps on the bottles.  Foreshortening Personal robots, when they are programmed to prepare or serve food, require more autonomy than robots in large-volume food service. A household robot might be programmed to prepare a meal of meat, vegetables, and beverages. The robot would ask questions such as these: • • • • • How many people will there be for this meal? Which type of meat is to be served? Which type of vegetable is to be served? How would you like the potatoes done? Or would you rather have rice? What beverages would you like? When all the answers were received, the robot would carry out the task of preparing the meal. The robot might also serve the meal, and then clean up the table and wash the dishes afterwards. See also PERSONAL ROBOT. FORESHORTENING In a robotic distance-measurement system, foreshortening is a false indication of the distance between a robot and a barrier, as measured along a specific straight-line path through three-dimensional (3-D) space. The phenomenon can occur when a barrier is oriented at a sharp angle with respect to the direction in which the range bearing is to be obtained. Sonar is particularly vulnerable to the problem, because it is difficult to focus acoustic waves into narrow beams. The illustration shows a dimensionally reduced example of how foreshortening can take place. The robot is shown as a shaded circle at left. Its direction of travel, and the favored direction (axis) of its sonar device, is directly from left to right (horizontally in this drawing). The sonar should ideally produce a range indication that is the same as the actual range, or the distance the robot must travel before it runs into the barrier. However, the field of view (FOV) of the sonar is 30°, or 15° to either side of the axis. The extreme right-hand edge of the sonar beam strikes the barrier before the central portion of the beam. Assuming the barrier has a surface sufficiently irregular to scatter the acoustic waves in all directions so the robot receives an echo from all portions of its sonar beam, the apparent range is significantly less than the actual range. The only solution to foreshortening problems of this sort is to minimize the FOV of the ranging equipment. In a work environment such as that shown in the drawing, the robot would be better off plotting a computer map of its surroundings, using a system more sophisticated than sonar. See also COMPUTER MAP, DISTANCE MEASUREMENT, FIELD OF VIEW (FOV), and RANGE PLOTTING.  Frame Actual range Apparent range Barrier Foreshortening FORWARD CHAINING A computer can act as a person knowledgeable in some field, such as engineering, weather forecasting, medicine, or even the stock market. Programs that make computers act like specialists are called expert systems. When running an expert system, you supply the computer with information, and the computer solves a problem based on that information. There are two ways in which the data can be supplied when using an expert system. You can input the facts one at a time, as the computer requests them; or you can input all the data at once, before the program begins working toward a solution. The latter method is forward chaining. The chain of reasoning starts from a single set of facts, and works forward until the problem is solved or a conclusion is reached. After a computer receives the data in a forward-chaining expert system, the inference engine uses rules, written in the software, to infer a solution or conclusion. If more information is necessary, the computer will let the operator know, usually by asking specific questions. Compare BACKWARD CHAINING. See also EXPERT SYSTEM. FRAME A frame is a mental symbol, a means of representing a set of things. Frames can be envisioned as “windows in the mind.” In artificial intelligence (AI), objects and processes can be categorized in frames. Suppose a robot is given the command, “Go to the kitchen and pour some water into a paper cup.” The robot goes through a series of deductions  Frankenstein Scenario concerning how to get this beverage, and how to obtain the object in which it is to be contained. First, the robot goes to the kitchen. Then it begins a search for the particular kind of beverage container that has been specified, in this case a paper cup. The illustration depicts this process. The first frame represents all the objects in the kitchen. Within this frame, a subframe is selected: eating and drinking utensils. Within this, the appropriate frame contains cups and tumblers; within this frame, the desired category is paper cups. Even this subset can be broken down further. One might specify 12-oz paper cups, white in color, designed to withstand hot beverages as well as cold. Eating and drinking utensils Things in kitchen Cups and tumblers Paper cups Frame Frames can apply to procedures as well as to the selection of objects. Once the robot has the proper utensil in its grasp, what is to be done? Did the robot’s user (human) want tap water, or is there some bottled water in the refrigerator? How about canned soda water? Maybe the user wants some of that mineral water she ran out of last week, in which case the robot must either come back and ask for further instructions, or else make a guess as to what substitute the user might accept. FRANKENSTEIN SCENARIO Science fiction is replete with stories in which some of the characters are robots or smart computers. Science-fiction robots are often androids. Such machines are invariably designed with the idea of helping humanity, although it often seems that the machines play roles in which some humans are “helped” at the expense of others. A recurring theme in science fiction involves the consequences of robots, or intelligent machines, turning against their makers, or coming  Function to logical conclusions intolerable to humanity. This theme is called the Frankenstein scenario, after the famous fictional android. A vivid example of the Frankenstein scenario is provided by the novel 2001: A Space Odyssey, in which Hal, an artificially intelligent computer on a space ship, tries to kill an astronaut. Hal somehow malfunctions, becomes paranoid, and believes that Dave, the astronaut, is intent on the computer’s destruction. Ironically, Hal’s paranoia brings about the very misfortune Hal dreads, because Dave is forced to disable Hal to save his own life. A machine might react logically to preserve its own existence when humans try to “pull the plug.” This could take the form of apparently hostile behavior, in which robot controllers collectively decide that humans must be eliminated. Because robots are supposed to preserve themselves according to Asimov’s three laws, a robotic survival instinct can be useful, but only up to a certain point. A robot must never harm a human being; that is another of Asimov’s laws. Another example of the Frankenstein scenario is the team of computers in Colossus: The Forbin Project. In this case, the machines have the best interests of humanity in mind. War, the computers decide, cannot be allowed. Humans, the computers conclude, require structure in their lives, and must therefore have all their behavior strictly regulated. The result is a totalitarian state run by a machine. See also ASIMOV’S THREE LAWS. FRONT LIGHTING In a robotic vision system, the term front lighting refers to illumination of objects in the work environment using a light source located at or near the robot’s own imaging sensors. The light from the source therefore reflects from the surfaces of the objects under observation before reaching the sensors. Because the location of the lamp is near the sensors, the robot sees minimal, or no, shadow effect in its work environment. Front lighting is used in situations where the surface details, particularly differences in color or shading, of observed objects are of interest or significance. For texture to show up, however, side lighting works best. Front lighting does not work particularly well in situations involving translucent or semitransparent objects, if their internal structure must be analyzed. Back lighting works best in these cases. Compare BACK LIGHTING and SIDE LIGHTING. FUNCTION A function is a mapping between set of objects or numbers. Functions are important in mathematics, and also in logic.  Function The drawing shows an example of a function as a mapping between two sets. Not all of the elements in the left-hand set (a few of which are shown by black dots) necessarily have counterparts in the right-hand set. Similarly, not all of the elements in the right-hand set (a few of which are shown by white dots) necessarily have counterparts in the left-hand set. If the mapping is to qualify as a function, it is possible for more than one element from the left-hand set to be mapped onto a single element in the right-hand set, but no element in the left-hand set can have more than one mate in the right-hand set. A function never maps a single element into more than one counterpart. Set mapped from Set mapped into Function As shown in the illustration, the set of all elements on the left that have mates in the right is called the domain of the function. The range of the function is the set of all elements on the right with corresponding elements in the set on the left. See also DOMAIN OF FUNCTION and RANGE OF FUNCTION. In logic, a function, more specifically called a logic function, is an operation that takes one or more input variables, such as X, Y, and Z, and generates a specific output for each combination of inputs. Logic functions are generally simpler than mathematical ones, because the input variables can only have two values: 0 (false) or 1 (true).  Futurist An example of a logic function in three variables is shown in the table. First, the logic AND operation is performed on X and Y. Then the logic OR operation is performed between (X AND Y) and the variable Z. Some logic functions have dozens of input variables; there is only one output value, however, for each combination of inputs. Function: example of a logic function X Y Z f (X, Y, Z) 0 0 0 0 0 0 1 1 0 0 1 1 1 1 1 1 0 0 1 1 0 1 0 1 0 1 0 1 0 1 1 1 Logic functions are important to engineers in the design of digital circuits, including computers. Often, there are several different possible combinations of logic gates that will generate a given logic function. The engineer’s job is to find the simplest and most efficient design. See also LOGIC GATE. The term function, or more specifically intended function, is often used in reference to the set of tasks or routines that a robotic device or controller is designed to perform or execute. This definition is completely independent of the mathematical and logical definitions. The mathematical equation, or set of equations, that represents a signal waveform is sometimes called a function. A function generator is a specialized circuit that generates waveforms whose curves are the graphs of specific mathematical functions. See GENERATOR. FUNCTION GENERATOR See GENERATOR. FUTURIST A futurist is a person who tries to predict, based on current technology and trends, what will be accomplished in a given scientific field in 5, 10, 50, 100, or more years. In robotics and artificial intelligence (AI), there is plenty of work for futurists.  Futurist Most futurists agree that robots will become more sophisticated, and more commonplace, as time goes by. There is some question as to exactly what form the robots will take. While it is fun to daydream about androids, these are often not the most practical and functional robots. There is theoretically unlimited potential in AI. In practice, however, things have moved more slowly than futurists of the twentieth century hoped. Reasoning processes are incredibly complex. Some futurists believe that all human thought processes can be broken down into interactions among particles of matter. If this is true, then it is technically possible (although difficult) to build a computer that is as smart as, or smarter than, a human being. Other scientists are convinced that human thought involves factors that cannot be defined or replicated in purely material terms. If this is the case, then a computer with superhuman intelligence might be impossible to build. Science-fiction authors have historically told stories about machines and scenarios, many of which have later become reality to a greater or lesser extent. For this reason, science-fiction writers have been called futurists.  G GANTRY ROBOT A gantry robot consists of a robot arm and end effector that employs threedimensional (3-D) Cartesian coordinate geometry for precise positioning. In one version of the gantry system, z-axis (up/down) movement is provided by a vertical shaft along which an assembly can slide. That assembly consists of a horizontal shaft, along which a horizontal arm at right angles to the shaft can slide on the y axis (forward/backward). A cable drive facilitates extension and retraction of the horizontal arm for x-axis (left/right) motion of the end effector. Gantry robots are used in industrial robotics to position end effectors over specific points on a horizontal plane surface. The end effector can be a gripper that picks up or releases objects, as in drop delivery. Alternatively, a rotating-shaft end effector can be used, as in a robot designed to tighten bolts. See also CABLE DRIVE, CARTESIAN COORDINATE GEOMETRY, END EFFECTOR, ROBOT ARM, X AXIS, Y AXIS, and Z AXIS. GAS STATION ROBOT Despite the rise in popularity of self-service gas stations, there are still people who would rather sit in their cars and have someone—or something—else do the dirty work. Robots are quite capable of filling your gas tank and washing your windshield. The drawing illustrates what a typical robotized drive-through filling station, or gas station robot, might look like. A person drives a car up to the paying station and inserts a credit card. This card has information concerning the make and year of the car, as well as credit account data. This tells the robot where it can find the gas-tank fill opening (right or left side of the car), and whether there are enough funds in the credit account to pay for a full tank of gas. Another method of car identification might  Gateway Overhead support Robot arm Gas hose Track for robot arm Paying station Driveway Gas station robot be the use of bar coding or a passive transponder, similar to the price tags on consumer merchandise. The robot must know the position of the car to within a millimeter or so. Otherwise, the nozzle might miss the fill pipe and spill gasoline on the pavement, or worse, put it in the car through the window. Object recognition helps prevent problems like this. Alternatively, a biased search can be used, letting the nozzle seek out the gas-tank fill opening. The opening itself is of a design such that the robot can open it and insert the nozzle, without any assistance from the human driving the car. Robotized filling stations, should they become the norm, will let people stay in their cars without getting dirty, cold, wet, or hot. The service of a well-designed, robotized gas station should be fast and efficient. Robots will have to be programmed not to “top off ” the gas tank to get a round number for the price. (This can result in overfilling the tank, and takes unnecessary time.) The robot will not forget to replace the gas cap, a perennial problem for some people who use “self-serve” gas stations. GATEWAY A gateway is a decision point in a specialized robotic-navigation process known as topological path planning. When a robot encounters a gateway, a decision must be made that affects the future path of the machine.  Generator An example of a gateway is an intersection between two streets. At a typical intersection where two straight roads cross each other at a right angle, a robotized vehicle can do any of four things: • • • • Continue straight ahead Turn left Turn right Backtrack When a mobile robot is programmed to travel from one point to another, gateways must frequently be dealt with. If the machine has a complete computer map of its work environment, and if the environment is not too complicated, every gateway possibility can reside in the controller’s memory or storage medium. If the work environment is complex, or if it changes with time, the decisions must be based on programming rather than brute-force data storage. See also COMPUTER MAP, RELATIONAL GRAPH, and TOPOLOGICAL PATH PLANNING. GENERATOR The term generator can refer to either of two devices. A signal generator is a source of alternating-current (AC) signal current, voltage, or power in an electronic circuit. An oscillator is a common example. An electric generator is a device that produces AC electricity from mechanical energy. Signal generator A signal generator is used for the purpose of testing audio-frequency (AF) or radio-frequency (RF) communications, detection, monitoring, security, navigation, and entertainment equipment. This includes various types of robotic sensing systems. In its simplest form, a signal generator consists of a simple electronic oscillator that produces a sine wave of a certain amplitude in microvolts ( V) or millivolts (mV), and a certain frequency in hertz (Hz), kilohertz (kHz), megahertz (MHz), or gigahertz (GHz). Some AF signal generators can produce several different types of waveforms, such as those shown in Fig. 1. The more sophisticated signal generators for RF testing have amplitude modulators and/or frequency modulators. A function generator is a signal generator that can produce specialized waveforms selected by the user. All electrical waveforms can be expressed as mathematical functions of time. For example, the instantaneous amplitude of a sine wave can be expressed in the form f (t) = a sin bt, where a is a constant that determines the peak amplitude and b is a constant that determines the frequency. Square waves, sawtooth waves, and all other periodic  Generator Sine Square Sawtooth Ramp Triangular Generator—Fig. 1 Irregular disturbances can be expressed as mathematical functions of time, although the functions are complicated in some cases. Most function generators can produce sine waves, sawtooth waves, and square waves. Some can also produce sequences of pulses. More sophisticated function generators that can create a large variety of different  Generator waveforms are used for testing purposes in the design, troubleshooting, and alignment of electronic apparatus. Electric generator An electric generator is constructed somewhat like a conventional electric motor, although it functions in the opposite sense. Some generators can also operate as motors; they are called motor/generators. Generators, like motors, are energy transducers of a special sort. A typical generator produces AC when a coil is rotated rapidly in a strong magnetic field. The magnetic field can be provided by a pair of permanent magnets (Fig. 2). The rotating shaft is driven by a gasolinepowered motor, a turbine, or some other source of mechanical energy. A commutator can be used with a generator to produce pulsating directcurrent (DC) output, which can be filtered to obtain pure DC for use with precision equipment. See also MOTOR. Coil Permanent magnet Permanent magnet N S N S Applied torque AC output Generator—Fig. 2  Global Positioning System (GPS) GLOBAL POSITIONING SYSTEM (GPS) The Global Positioning System (GPS) is a network of wireless location and navigation apparatus that operates on a worldwide basis. The GPS employs several satellites, and allows determination of latitude, longitude, and altitude. It is used in some mobile robotic systems for guidance when extreme, localized precision is not necessary. All GPS satellites transmit signals in the ultra-high-frequency (UHF) radio spectrum. The signals are modulated with codes that contain timing information used by the receiving apparatus to make measurements. A GPS receiver determines its location by measuring the distances to four or more different satellites, and using a computer to process the information received from the satellites. From this information the receiver can give the user an indication of position accurate to within a few meters. See also DISTANCE MEASUREMENT. GRACEFUL DEGRADATION When a portion of a computer system malfunctions, it is desirable to have the computer keep working even if the efficiency is impaired. If a single component causes the whole computer to fail, it is called a catastrophic failure. This can generally be prevented by good engineering, including the use of backup systems. In graceful degradation, as the number of component failures increases, the efficiency and/or speed of the system gradually Relative system efficiency Increasing component failures Graceful degradation  Graphical Path Planning declines, but does not instantly drop to zero. The illustration is a graph of the behavior of a hypothetical robotic system with graceful degradation. In the event of a subsystem malfunction, a sophisticated computer or robot controller can use other circuits to accomplish the tasks of the failed part of the system temporarily. The human operator or attendant is notified that something is wrong, and technicians can fix it, often with little or no downtime. Compare FAULT RESILIENCE. GRAPHICAL PATH PLANNING Graphical path planning is a method of navigation used by mobile robots. It is a specialized scheme or set of schemes for the execution of metric path planning. In graphical path planning, all possible routes are plotted on a computer map of the work environment. These routes can be chosen in various ways, by employing specific algorithms. In an open work environment (that is, one in which there are no hazards or obstructions), the best routes are usually straight lines between the nodes, or stopping points (Fig. 1). The algorithm for determining these paths is comparatively simple; it can be represented by a set of linear equations in the robot controller. An obstacle, barrier, or hazard can complicate this scenario, but only if it intersects, or nearly intersects, one of the lines determined by the linear equations. To avoid mishaps, the algorithm can be modified to include a statement to the effect that the machine must never come closer than a certain distance to an obstacle, barrier, or hazard. Proximity sensing can be employed to detect these situations. In a work environment in which there are numerous obstacles or hazards, or where there are barriers such as walls separating rooms and hallways, the straight-line algorithm is not satisfactory, even in amended form, because too many modifications are necessary. One scheme that works quite well in this type of environment is the Voronoi graph. The Graphical path planning—Fig. 1  Grasping Planning paths are defined as sets of points at the greatest possible distances from obstacles, barriers, or hazards. In a hallway, for example, the path goes down the middle. The same is true as the robot passes through doorways. The paths in other places depend on the locations of the nodes, and the arrangement of obstructions in the rooms or open areas (Fig. 2). See also COMPUTER MAP. Compare METRIC PATH PLANNING and TOPOLOGICAL PATH PLANNING. Graphical path planning—Fig. 2 GRASPING PLANNING Grasping planning refers to the scheme that a robot arm and gripper use to get hold of a chosen object. Suppose a person tells a robot to go to the kitchen and get a spoon. The robot uses gross motion planning to find the kitchen, and fine motion planning to locate the correct drawer and determine which objects in the drawer are the spoons. Then the gripper must grasp a spoon, preferably by the handle rather than by the eating end. The robot must not get a fork, or two spoons, or a spoon along with something else such as a can opener. Hopefully, the silverware is arranged logically in the drawer, so spoons are not randomly mixed up with forks, knives, can openers, and other utensils. This can be ensured by programming, as long as the robot (but only the robot) has access to the drawer. If there are children in the household, and if they get into the silverware drawer, the robot had better be able to cope with mixed-up utensils. Then, getting a spoon becomes a form of bin picking problem. Close-up, detailed machine vision, such as an eye-in-hand system, can ensure that the gripper gets the right utensil in the right way. Tactile sensing might also be used, because a spoon “feels” different than any other kind of utensil. See also BIN PICKING PROBLEM, EYE-IN-HAND SYSTEM, FINE MOTION PLANNING, GROSS MOTION PLANNING, OBJECT RECOGNITION, and TACTILE SENSING.  Gross Motion Planning GRAVITY LOADING Gravity loading is a phenomenon that introduces positioning error into robot arms as a result of the force of gravity. All robot arms are comprised of materials that bend or stretch to some extent; no known substance is perfectly rigid. In addition, all materials have some mass; thus, in a gravitational field they also have weight. The weight of the robot arm and end effector always causes some bending and/or stretching of the materials from which the assembly is made. The effect can be exceedingly small, as in a telescoping, vertically oriented robot arm; or it can be larger, as in a long, jointed robotic arm. However, the effect is never entirely absent. The error caused by gravity loading is not always significant. In situations where gravity loading causes significant positioning errors, a scheme for correction is necessary. See also ERROR CORRECTION. GRAYSCALE Grayscale is a method of creating and displaying digital video images. As its name implies, a grayscale vision system is color-blind. Each image is made up of pixels. One pixel is a single picture (pix) element. The pixels are tiny squares, each with a shade of gray that is assigned a digital code. There are three commonly used schemes for rendering pixels in grayscale: percentage-of-black, 16 shades of gray, and 256 shades of gray. In the percentage scheme, there are usually 11 levels according to the following sequence: {black, 90 percent black, 80 percent black, …, 20 percent black, 10 percent black, white}. Sometimes the brightness is broken down further, into increments of 5 percent or even 1 percent rather than 10 percent; such gradations tend to be imprecise because computer digital codes are binary (power-of-2), not decimal (power-of-10). In the 16-shade scheme, four binary digits, or bits, are needed to represent each level of brightness from black = 0000 to white = 1111. In the 256-shade scheme, eight binary digits are used, from black = 00000000 to white = 11111111. See also COLOR SENSING and VISION SYSTEM. GRIPPER See ROBOT GRIPPER. GROSS MOTION PLANNING Gross motion planning is the scheme a mobile robot employs to navigate in its work environment without bumping into things, falling down stairs, or tipping over. The term can also refer to the general, programmed  Groundskeeping Robot sequence of movements that a robot arm undergoes in an industrial robotic system. Gross motion planning can be done using a computer map of the environment. This tells it where tables, chairs, furniture, and other obstructions are located, and how they are oriented. Another method is to use proximity sensing or a vision system. These devices can work in environments unfamiliar to a robot, and for which it has no computer map. Still another method is the use of beacons. Suppose a personal robot is told to go to the kitchen and get an apple from a basket on a table. The robot can employ gross motion planning to scan its computer map and locate the kitchen. Within the kitchen, it needs some way to determine where the table is located. Finding the basket, and picking an apple from it (especially if there are other types of fruit in the basket, too), requires fine motion planning. Compare FINE MOTION PLANNING and GRASPING PLANNING. GROUNDSKEEPING ROBOT There are plenty of jobs for personal robots in the yard around the house, as well as inside the house. Two obvious applications for a groundskeeping robot includes mowing the lawn and removing snow. In addition, such a machine might water and weed a garden. Riding mowers and riding snow blowers are easy for sophisticated mobile robots to use. The robot need not be a biped; it needs only to have a form suitable for riding the machine and operating the controls. Alternatively, lawn mowers or snow blowers can be robotic devices, designed with that one task in mind. The main challenge, once a lawn-mowing or snow-blowing robot has begun its work, is to do its work everywhere it is supposed to, but nowhere else. A robot owner does not want the lawn mower in the garden, and there is no point in blowing snow from the lawn (usually). Such a robot should therefore be an automated guided vehicle (AGV). Currentcarrying wires can be buried around the perimeter of your yard, and along the edges of the driveway and walkways, establishing the boundaries within which the robot must work. Inside the work area, edge detection can be used to follow the line between mown and unmown grass, or between cleared and uncleared pavement. This line is easily discernible because of differences in brightness and/or color. Alternatively, a computer map can be used, and the robot can sweep along controlled and programmed strips with mathematical precision. The hardware already exists for groundskeeping robots to withstand all temperatures commonly encountered in both summer and winter, from Alaska to Death Valley. Software is more than sophisticated enough for  Gyroscope ordinary yard-maintenance and snow-removal tasks. The only challenge remaining is to bring the cost down to the point that the average consumer can afford the robot. See also AUTOMATED GUIDED VEHICLE, COMPUTER MAP, EDGE DETECTION, and PERSONAL ROBOT. GUIDANCE SYSTEM In robotics, guidance system refers to the hardware and software that lets a robot find its way in its work environment. In particular, it refers to gross motion. For detailed information, see AUTOMATED GUIDED VEHICLE, BEACON, BIASED SEARCH, COMPUTER MAP, DIRECTION FINDING, DIRECTION RESOLUTION, DISTANCE MEASUREMENT, DISTANCE RESOLUTION, EDGE DETECTION, EMBEDDED PATH, EPIPOLAR NAVIGATION, GLOBAL POSITIONING SYSTEM (GPS), GROSS MOTION PLANNING, GYROSCOPE, LOG POLAR NAVIGATION, OBJECT RECOGNITION, PARALLAX, PROXIMITY SENSING, RADAR, SONAR, and VISION SYSTEM. GYROSCOPE A gyroscope or gyro is a device that is useful in robot navigation. It forms the heart of an inertial guidance system, operating on the basis of the fact that a rotating, heavy disk tends to maintain its orientation in space. Bearings Rotating disk Bearings Gyroscope  Gyroscope The illustration shows the construction of a simple gyroscope. The disk, made of massive material such as solid steel or tungsten, is mounted in a gimbal, which is a set of bearings that allows the disk to turn up and down or from side to side; conversely, the bearings allow the entire assembly (except for the disk) to undergo pitch, roll, and yaw while the disk remains fixed in its spatial orientation. The disk is usually driven by an electric motor. A gyroscope can be employed to keep track of a robot’s direction of travel, or bearing, in three-dimensional (3-D) space without reliance on external objects, beacons, or force fields. Gyroscopes allow the accurate operation of guidance systems for a limited time, because they tend to change their orientation slowly over long periods. In addition, gyroscopes are susceptible to misalignment in the event of physical shock. See also PITCH, ROLL, and YAW.  H “HACKER” PROGRAM One of the earliest experiments with artificial intelligence (AI) was done with an imaginary robot, entirely contained within the “mind” of a computer. A student named Gerry Sussman wrote a program called “Hacker,” in a computer language known as LISP. The result was a little universe in which a robot could stack blocks on each other. Sussman created laws of physics in the imaginary universe. Among them were things such as • • • • • • Blocks X, Y, and Z each weigh 5 lb. Blocks V and W each weigh 50 lb. The robot can lift no more than 10 lb. Only one object can occupy a given space at a given time. The robot knows how many blocks there are. The robot can find blocks if they are not in direct sight. Illustration 1 shows the five blocks lying around, as they might appear on the computer monitor, along with the robot. Sussman gave commands to the robot, such as, “Stack the blocks all up, one on top of the other.” As stated, this command is impossible, because it requires the robot to lift a block weighing 50 lb (either V or W), and the robot is capable of lifting only 10 lb (see illustration 2). What would happen? Would the robot try forever to lift a block beyond its limit of strength? Or would it tell Sussman something like, “Unable to do this”? Would it go after either block V or W first, trying to get it on top of one of the lighter blocks, or on top of the other heavy block? Would it pick up all the lighter blocks X, Y, and Z in some sequence, stacking them vertically on top of V or W? Would it put two light blocks on V, and the remaining light block on W, and then give up? Eventually, the robot would run into the impossibility of the command. But how long would it try, and what would it try, before quitting?  Hallucination Z Y V X W Z Y X V 2. Now what? W 1. Why is X trying to hide? X V 3. This is all right. “Hacker” program Y Z W Y X V 4. Is this all right too? Z W Another command might be, “Stack the blocks so that lighter ones are on top of heavier ones.” This can be done according to the rules written above. But there exist several different possible ways (two of these are shown in illustrations 3 and 4). Would the robot hesitate, unable to make a decision? Or would it go ahead and accomplish the task in some way? If the experiment were repeated, would the result always be the same, or would the robot solve the problem a different way each time? Numerous AI researchers have written programs similar to “Hacker,” creating “computer universes” in an attempt to get machines to think and learn. The results have often been fascinating and unexpected. HALLUCINATION In a human being, a hallucination occurs when the senses deliver phantom messages. This can happen in mental illness, or under the influence of  Handshaking certain drugs. Hallucinations can be, and often are, combined with delusions, or misinterpretations of reality. An example is the person who thinks that spies are after him or her, and who sees sinister figures lurking behind trees or in dark alleyways. Sophisticated computers can appear to have hallucinations and delusions. The likelihood of such malfunctions, taking place in bizarre and often inexplicable ways, increases as systems become more complex. This is because, as computers become smarter, the number of components, pathways, and nodes increases in exponential proportion, and the probability of a component failure or stray signal thus “blows up.” Computer components are, in general, exceptionally reliable; however, given great enough numbers of them, strange things can happen, and have happened. Experienced personal computer users and technicians know this. Some researchers in artificial intelligence (AI) believe that electronic hallucinations or delusions might someday result from improper design and care of machines. These researchers suspect that machines, as they evolve and become more intelligent, can develop “hangups,” just as people do. At present, malicious human operators cause more problems directly, by means of such schemes as hacking and the writing of computer viruses, than “computers gone mad.” In a few decades, however, autonomous robots might become able to program and maintain themselves to a large extent, and the situation might change. See AUTONOMOUS ROBOT. HAND See ROBOT GRIPPER. HANDSHAKING In a digital communications system, accuracy can be optimized by having the receiver verify that it has received the data correctly. This is done periodically—say, every three characters—by means of a process called handshaking. The process goes as follows, as illustrated in the figure. First, the transmitter sends three characters of data. Then it pauses, and awaits a signal from the receiver that says either of the following: • (a) All three characters have familiar formats. • (b) One or more characters has an unfamiliar format. If the return signal is (a), the transmitter sends the next three characters. If the return signal is (b), the transmitter repeats the three characters. In computer systems, the term handshaking refers to a method of controlling, or synchronizing, the flow of serial data between or among devices. The synchronization is accomplished by means of a control wire in  Hard Wiring Send 3 characters Return signal: retransmit Signal received Return signal: send next No Format proper? Yes Handshaking hardware, or a control code in the programming. Hardware handshaking is used when direct wire or cable links are possible, such as between a personal computer and a serial printer. Software handshaking is similar to the process used for communications systems. HARD WIRING In a computer or autonomous robot, the term hard wiring refers to functions that are built directly into the machine hardware. Hard wiring cannot be changed without rearranging physical components, or changing the interconnecting wires. Sometimes, the expression firmware is used to mean hard wiring, although technically that is a misuse of the term. An ideal computer (that is, a computer with infinite processing power and zero error rate, which can exist only in theory) could be programmed to do anything without having to move a single physical component. Of course, the components must be hooked up together somehow, but in the ideal case, functions could be changed just by reprogramming the machine. This has been realized to a large extent in recent years by the use of high-speed, high-capacity data storage media. Hard wiring does have some advantages over software control. Most significant is the fact that hard-wired functions can be done at a higher  Heuristic Knowledge rate of speed than processes that require access to mechanical storage media. However, as nonvolatile storage media without moving parts become more widely available, this advantage of hard wiring will gradually erode. Compare FIRMWARE. HERTZ Hertz, abbreviated Hz, is the fundamental measure of alternating-current (AC) frequency. A frequency of 1 Hz is equivalent to one cycle per second. In fact, the word “hertz” is interchangeable with the expression “cycles per second.” Frequency is often expressed in units of kilohertz (kHz), megahertz (MHz), and gigahertz (GHz). A frequency of 1 kHz is equal to 1000 Hz; a frequency of 1 MHz is equal to 1000 kHz or 106 Hz; a frequency of 1 GHz is equal to 1000 MHz or 109 Hz. The speed at which digital computers operate is often specified in terms of frequency. The higher the frequency, the faster a microprocessor can work, and the more powerful can be the computer that uses the chip— if all other factors remain constant. The reason that higher frequency translates into a more powerful chip is simply that, as the frequency increases, more and more instructions can be executed, and thus more operations done, per unit time. The clock frequency of the microprocessor is, however, only one of several factors that determine the processing speed of a computer. HEURISTIC KNOWLEDGE Can computers and robots learn from their mistakes, and improve their knowledge by trial and error? Is it possible for a machine, or a network of machines, to evolve on its own? Some artificial intelligence (AI) researchers believe so. The existence of heuristic knowledge, or the ability of a machine to become smarter based on its real-world experience— literally learning from its own mistakes—is a classical characteristic of true AI. Suppose a powerful computer is developed that can evolve to higher and higher levels of knowledge. Imagine that, one day after the machine has been put into operation, it has intelligence equivalent to that of a 10-yearold human; and after two days, it is as smart (in a rudimentary sense) as a 20-year-old. Suppose that after three days, the machine has knowledge equivalent to that of a 30-year-old research engineer. Suppose that more and more memory is added, so that the limit of knowledge is determined only by the speed of the microprocessor. What will such a computer be like after a month? Will it have the knowledge of a 300-year-old person (if people lived that long)? Moreover, does an ever-increasing level of intelligence imply that a machine can also become “wise”?  Hexadecimal Number System Machine knowledge becomes far more powerful when computers are given the ability to control mechanical devices, as is the case with autonomous robots. Intelligence and knowledge alone cannot build cars, bridges, aircraft, and rockets. Perhaps dolphins are as smart as people, but these marine mammals lack hands and fingers with which to manipulate things. A computerized robot is to a computer as a human being is to a dolphin. Can computers ever become smarter than, and perhaps more powerful than, their makers? Some scientists are concerned that AI will be misused, or that it could evolve on its own with unintended, unexpected, and unpleasant results. Other researchers believe that the potential benefits of ever-increasing machine knowledge will always outweigh the potential dangers, and that we can always pull the plug if things get out of control. HEXADECIMAL NUMBER SYSTEM See NUMERATION. HIERARCHICAL PARADIGM The term hierarchical paradigm refers to the oldest of three major approaches to robot programming. A robot that employs the hierarchical paradigm relies largely on advance planning to carry out its assigned tasks. In the most sophisticated robot systems, there are three basic functions, known as plan/sense/act. The hierarchical paradigm simplifies this to plan/act. The original idea for this paradigm was based on an attempt to get a smart robot to mimic human thought processes. The robot first senses the nature of its work environment, plans an action or sequence of actions, and then carries out those actions. In some systems this process occurs only once, at the beginning of the task; in other systems the planning step is repeated at intervals during the execution of the task. The hierarchical paradigm has also been called the deliberative paradigm, because of its reliance on creating fixed models of the work environment. The robot controller functions in a sense as if it is “cogitating” or “deliberating” a strategy prior to carrying it out. This scheme has proven too simplistic for many practical scenarios, and around the year 1990, it was superseded by more advanced programming methods. Compare HYBRID DELIBERATIVE/REACTIVE PARADIGM and REACTIVE PARADIGM. HIGH-LEVEL LANGUAGE The term high-level language refers to programming languages used by humans in their interactions with computers. The various high-level languages each have advantages in some types of work, and shortcomings in others.  Hold High-level language consists of statements in English (or some other written human language). This allows people to work with computers on a sort of conversational level. Most students find high-level languages easy to learn. The best way to learn these languages is to “play computer,” thinking strictly by rules of logic. Because of the pure logic in programming, computers might someday be used to develop new programs for other computers. Compare MACHINE LANGUAGE. See also HEURISTIC KNOWLEDGE. HOBBY ROBOT A hobby robot is a robot intended mainly for amusement and experimentation. Such a machine is usually autonomous, and contains its own controller. It is, in effect, a sophisticated toy. Hobby robots often take humanoid form; these are androids. They can be programmed to give lectures, operate elevators, and even play musical instruments. Wheel drives are commonly used rather than bipedal (twolegged) designs, because wheels work better than legs, are easier to design, and cost less. However, some hobby robots are propelled by track drives; others have four or six legs. Some hobby robots are adaptations of industrial robots. Robot arms can be attached to a main body. Vision systems can be installed in the robot’s head, which can be equipped to turn to the right and left, and to nod up and down. Speech recognition and speech synthesis can allow a hobby robot to converse with its owner in plain language, rather than by means of a keyboard and monitor. This makes the machine much more human-like and user-friendly. Perhaps the most important feature for a hobby robot is artificial intelligence (AI). The “smarter” the robot, the more fun it is to have around. It is especially interesting if a machine can learn from its mistakes, or be taught things by its owner. Hobby robot societies exist in the United States and several other developed countries. They evolve and change their names often. If you live in a large city, you might be near such an organization. See also PERSONAL ROBOT. HOLD Hold, also called holding, is a condition in which the movements of part, or all, of a robot manipulator are temporarily brought to a halt. When this occurs, braking power is maintained, so the halted parts resist movement if outside pressure is applied. Common methods of ensuring braking force involve the use of a hydraulic drive or a stepper motor.  Home Position Holding can be a part of the programmed movement sequence for a robot arm and end effector. A good example is a situation in which a gantry robot is used to position a component for drop delivery. See also DROP DELIVERY, GANTRY ROBOT, HYDRAULIC DRIVE, and STEPPER MOTOR. HOME POSITION In a robot manipulator, the home position is a point at which the end effector normally comes to rest. When the robot is shut down, or when it must be reset, the machine reverts to its home position. When a coordinate system is used to define the location of the end effector, the home position is often assigned to the origin point. Thus, for example, in a robot arm and end effector using two-dimensional (2-D) Cartesian coordinate geometry, the home position can be assigned the value (x, y) = (0, 0). HOUSEHOLD ROBOT See PERSONAL ROBOT. HUMAN ENGINEERING Human engineering refers to the art of making machines, especially computers and robots, easy to use. This is sometimes also called user-friendliness. A user-friendly computer program allows the machine to be operated by someone who knows nothing about computers. Bank automatic-teller machines (ATMs) are a good example of devices that employ userfriendly programming. Increasingly, libraries are computerizing their card catalogs, and it is important that the programs be user-friendly so that people can find the books they want. There are many other examples. A user-friendly robot can carry out orders efficiently, reliably, and reasonably fast. Ideally, a human operator can say something like, “Go to the kitchen and get me an apple,” and (assuming there are any apples in the kitchen) the robot will return in a minute or two, holding an apple. A seemingly simple task like this is difficult to program into a machine, as researchers have found out. Even the most basic tasks are complex in terms of the number and combination of digital logic operations. One of the most important considerations in human engineering is artificial intelligence (AI). It is much easier to communicate with a machine that is “smart,” compared with one that is “stupid.” It is especially enjoyable if the machine can learn from its mistakes, or show ability to reason. Speech recognition and speech synthesis also help make computers and robots user-friendly. See also HEURISTIC KNOWLEDGE, SPEECH RECOGNITION, and SPEECH SYNTHESIS.  Hybrid Deliberative/Reactive Paradigm HUMANOID ROBOT See ANDROID. HUNTING Hunting is the result of overcompensation in a servomechanism. It is especially likely when there is not enough hysteresis, or sluggishness, in the system response. Any circuit or device designed to lock onto something, by means of error correction, is subject to hunting. It takes the form of a back-and-forth oscillation between two conditions. If severe, it can go on indefinitely. In less serious cases, the system eventually settles on the correct level or position (see the illustration). Error Hunting Hunting is eliminated by careful design of feedback systems, so there is just the right amount of hysteresis. See HYSTERESIS LOOP and SERVOMECHANISM. HYBRID DELIBERATIVE/REACTIVE PARADIGM The hybrid deliberative/reactive paradigm is an approach to smart-robot programming that combines the attributes of two simpler schemes, known as the hierarchical paradigm and the reactive paradigm. The hybrid paradigm came into favor during the 1990s. It operates according to the principle plan/sense/act. The actions are based on advance planning and also on the outputs of sensors from moment to moment. Before the task in begun, the robot generates a work plan. This is known as mission planning, and is a form of deliberation. A complex task Left Right Time  Hydraulic Drive is broken down into several components, or subtasks. Each subtask has its own subplan. Once the robot has begun executing the job, it carries out the plan and the subplans subject to modifications that may be necessary as the work environment changes. These changes are the results of signals from the sensors. In a typical robot that uses the hybrid paradigm, deliberations occur at intervals of several seconds, while reactions take place at a rate of many times per second. Compare HIERARCHICAL PARADIGM and REACTIVE PARADIGM. HYDRAULIC DRIVE A hydraulic drive is a method of providing movement to a robot manipulator. It uses a special hydraulic fluid, usually oil-based, to transfer forces to various joints, telescoping sections, and end effectors. The hydraulic drive consists of a power supply, one or more motors, a set of pistons and valves, and a feedback loop. The valves and pistons control the movement of the hydraulic fluid. Because the hydraulic fluid is practically incompressible, it is possible to generate large mechanical forces over small surface areas, or, conversely, to position large-area pistons with extreme accuracy. The feedback loop consists of one or more force sensors that provide error correction and ensure that the manipulator follows its intended path. Hydraulically driven manipulators are used when motions must be rapid, precise, and repeated numerous times. Hydraulic systems are also noted for the ability to impart considerable force, so they are good for applications involving heavy lifting or the application of large amounts of pressure or torque. In addition, hydraulically driven robot manipulators resist unwanted movement in the presence of external forces. Compare PNEUMATIC DRIVE. HYSTERESIS LOOP A hysteresis loop (the word is pronounced “his-ta-REE-sis”) is a graph that shows the sluggishness of response in a servomechanism. The illustration shows a hysteresis loop for a typical thermostat, used for control of the indoor air temperature in a house. The horizontal scale shows the room temperature in degrees Celsius (°C). On/off conditions for heating and cooling are shown on the vertical scales. Notice that there is a small range of temperatures, from about 18.5°C to 21.5°C, within which the temperature fluctuates. This prevents the system from rapidly oscillating back and forth between heating and cooling states, but it is a narrow enough temperature range so that the people in the room don’t get too hot or cold.  Hysteresis Loop On Off Heating Off On 18 20 Temperature, Degrees Celsius 22 Hysteresis loop All servomechanisms employ feedback of some kind. There must always be some hysteresis built into the feedback response. This hysteresis is often a natural result of the environment; for example, it takes some time for the temperature in a house to warm up and cool down by 3°C as depicted in the illustration. However, in a tiny temperature-regulated chamber intended to ensure the stable operation of the controller in a robot working in an extreme environment such as outer space, the hysteresis must be incorporated into the electronic design of the feedback circuit or thermostat. Otherwise, overreactions can be so severe that the system constantly cycles between states. See also SERVOMECHANISM.  Cooling This page intentionally left blank. I IF/THEN/ELSE In computers and smart robots, choices must often be made in the execution of a program. One of the most common programming decision processes is called IF/THEN/ELSE. It can be expressed as a sentence: “If A, then B; otherwise (or else) C.” An example of an IF/THEN/ELSE process is shown in the illustration. The intent is to determine the absolute value of a real number. Suppose a computer is working with an input number, designated x. If x is negative (that is, if x 0), then x must be multiplied by 1 to obtain the absolute value |x|. If x is zero or positive, then x is equal to its absolute value. The computer must compare the numerical value of x with zero. The machine will then output the absolute value of the number, by either multiplying x by 1 or by leaving x alone. IF/THEN/ELSE processes are especially useful command structures for robots. You might tell a robot, “Go to the kitchen and get me a paper napkin.” The robot controller has a command structure stored on its hard drive or in memory. It needs an alternative in case there are no paper napkins in the kitchen. The programming might take the form: “If this command can be executed, then carry out the task. Otherwise, output the audio statement, ‘Your order cannot be completed because there are no paper napkins in the kitchen.’ ”  Ignorant Coexistence Input x No x < x =" 39" y =" 75" z =" 51" z =" 51" speed =" 17" y =" 75" speed =" 25" x =" 39" speed =" 13" kbps =" 1000" mbps =" 1000" 53816 =" 2" 94 =" (4" 1011110 =" (0" 2 =" 1." r =" 1." violet =" 0–17%;" blue =" 18–33%;" green =" 34–50%;" yellow =" 51–67%;" orange =" 68–83%;" red =" 84–100%." time =" t" 92 =" 846,400" height =" 1" width =" 1" depth =" 1" height =" 2" width =" 2" depth =" 2" 102 =" 100." 103 =" 1000" volume =" 8" volume ="1" area =" 1" area =" 4" khz =" 1000" khz =" 1000" d=" h" d=" In" laser =" Sensor" y =" f" z =" f">