The Myths
Once upon a time a sculptor named Pygmalion carved a statue of a girl. It was so very beautiful and lifelike that it did indeed come to life; he named her Galatea and they lived happily ever after and inspired a play and a musical comedy: "Pygmalion" and "My Fair Lady." For as long as history and archaeology tell us about, people have been carving magic idols, believing them to be sufficiently alive to hear and obey the prayers of their makers. ("Magic" is the control of nature without the restrictions of what killjoy scientists call "laws of nature."). Animism is the most primitive form of religion; it attributes life and will to "the spirits of" trees and rocks and other objects. To this day some have the feeling that machines of metal and plastic have life and will. Have you, yourself, gentle reader, ever been angry with a car or a TV and struck or kicked it to punish its recalcitrance? A history of the development of a new computer was actually entitled "The Soul of a New Machine." In the early days of the SEAC computer at the National Bureau of Standards, a famous programmer, after a long period of frustration, went into the computer room and had a long talk with the machine in which she sincerely begged for a greater degree of cooperation. (I was told this by a friend of mine, an engineer who worked with her.)
The Fiction
The word "robot" was coined by a Czeck playwright, Karel Capek, in the early 1920's. He wrote a play called "R.U.R.: Rossums's Universal Robots." In it he used the word robot which he adapted from the Czeck word for work. The play was an ordinary Sci-Fi opus in which the mad scientist invents artificial people, robots, lacking only souls, to do the world's work. The robots rebel, but the good guys win. The play died but the word lived on and is now the same in all languages. Science fiction also lived on and there have been innumerable similar scripts for books and plays and movies. Even the ballet, Coppelia, is about a robot girl with whom the living hero falls in love; Pinocchio is a living wooden boy. Another literary robot was the monster created by imaginary Dr. Frankenstein who was created by real Mary Shelley. The good doctor modestly re-assembled existing human parts instead of starting from scratch.
Workers
Most people dislike doing most work and find a variety of ways to avoid it. Some simply shirk; others engineer machines to do it for them. The earliest solution to the problem was to force somebody else to do the work. The practice was called slavery and it was used all over the world. There were a variety of rules dealing with ethnic origin, capture in conquest, people imprisoned for crime (mostly political), and the like. It worked pretty well for the slave owners and still does in some corners of the world. Hitler and Stalin were the most recent large scale European practitioners. It is largely out of fashion at present and, in the United States at least, laws against peonage tend to limit equivalent practices. Some slave language lives on in the language of smart machines: MASTER, SLAVE, COMMAND, OBEY, SERVOMECHANISM, SERVO. Slavery fell out of favor in the western world in the 19th century and was replaced by hiring people for pay, in money and keep. This practice is called employment and is now used all over the world. A variety of rules vary from only a legalistic distinction from slavery to a very favorable treatment indeed of the hired person. If the hired person does domestic work he or she is, or was, called a servant and if the hired person works in a business he or she is called an employee. There are several objections to getting work done by employees. Human workers are not always energetic, reliable, docile, smart, and easily led. They are not always cheap, and those with the desired skills are not always available.
Artificial People
So, for as long as recorded history, people have wanted to make real artificial people to be their slaves. In the middle ages, when clockwork mechanisms were developed, articulated models of people were engineered, powered by springs or descending weights. There was even a mechanical man who wrote a few Chinese characters on paper, made as a gift to the emperor of China. A mechanical man, built in 1770, is still in working order in the Swiss Musee d'Art et d'Histoire. Town clocks exhibiting a parade of animated characters (JACKWORK) have been refurbished as tourist attractions in Europe and a brand new one has been built and installed near the Pompidou Center in Paris; I have watched a crowd of people enthralled by its gyrations. Disney World has lifelike machines acting out scenes from American history as well as other JACKWORK displays. Every toy store has mechanical dolls. The ultimate dream was, and is, a mechanical person who is the slave of a real person. This is pretty easy to accomplish in the imagination and became one of the staple themes of science fiction (one of the greatest oxymorons of all time.) Not quite so easy for real. The emotional appeal of robots approaches that of religion and patriotism. In 1986 I visited a research institute in China and saw some primitive robot engineering. When I said that China had so much cheap labor that there was no economic benefit from robots, I was answered, passionately, " China needs robots!" For prestige, perhaps, as with national airlines subsidized by poor countries to fly international routes. After all, the United States sent men to the moon for competitive prestige after the Russians orbited Uri Gagarin for competitive prestige. In contrast to the alchemy-like attempts by some at artificial people as mechanical slaves, others were content to engineer real machines intended merely to be machines and to do real work. These more modest folk were much more successful, since they did not have the self imposed requirement of anthropomorphism (man shape) and merely designed machines to suit their jobs. It has became useful for commercial prestige to apply the words "robot" and "robotic" to any smart machine.
Real Robots
After the technology explosion during World War II, George Devol, a successful inventor and entrepreneur, met engineer Joseph Engelberger at a cocktail party in Connecticut. They were both bright, enthusiastic, and imaginative. Together they made a serious and commercially successful effort to develop a real, working robot. Their concept was to use an electronic digital computer as the "brain" (i.e. program controller) and servomechanisms as the "muscles." A mechanism was designed suggestive of a human arm. (Anthropomorphism remained an article of faith.) It resembled a tank turret and cannon with additional axes of rotary motion at the outboard end of the cannon (the "wrist"). Servo actuators used hydraulic cylinders, electro-hydraulic servo valves, and electronic encoders for feedback. The controller was the digital computer. The machine was trade named the "Unimate." The hard part of entrepreneuring is funding. They persuaded Norman Schafler of Condec Corporation in Danbury that they had the basis of a commercial success, and Unimation, Inc. was born. Ultimately Unimation was acquired by Westinghouse and the entrepreneurs' dream of wealth was achieved. Most Unimates were sold to extract die castings from die casting machines and to perform spot welding on auto bodies, both tasks being particularly hateful jobs for people. Both applications were commercially successful, i.e., the robots worked reliably and saved money by replacing people. An industry was spawned and a variety of other tasks were also performed by robots, such as loading and unloading machine tools. The robot idea was hyped to the skies and became high fashion in the Boardroom. Presidents of large corporations bought them, for about $100,000 each, just to put into laboratories to "see what they could do;" in fact these sales constituted a large part of the robot market. Some companies even reduced their ROI (Return On Investment criteria for investment) for robots to encourage their use. The image of the "electronic brain" as the principal part of the robot was pervasive. Computer scientists were put in charge of robot departments of robot customers and of factories of robot makers. Many of these people knew little about machinery or manufacturing but assumed that they did. (There is a common delusion of electrical engineers that mechanical phenomena are simple because they are visible. Variable friction, the effects of burrs, minimum and redundant constraints, non-linearities, variations in workpieces, accommodation to hostile environments and hostile people, etc. are like the "Purloined Letter" in Poe's story, right in front of the eye, yet unseen.) They also had little training in the industrial engineer's realm of material handling, manufacturing processes, manufacturing economics and human behavior in factories. As a result, many of the experimental tasks in those laboratories were made to fit their robot's capabilities but had little to do with the real tasks of the factory. My own robot company (MOBOT Corp.) was once visited by a computer science Ph.D. He had been in charge of robot applications in a major American manufacturing company (a customer of mine) and was now president of a competing robot company. After the factory tour he said, quite sincerely, that our very large machines were unnecessary because they could be replaced by integrated circuit chips.
Anthropomorphism
Robots were described in anatomical language: they had arms, wrists, hands, fingers, and, of course, brains. Better mechanism designs which were not analogous to humans (not anthropomorphic) were denounced by zealots as "not real robots." All animals have hinged joints, like elbows, and none have sliding parts, like trombones. Bugs, fish, puppy dogs, people. Probably something to do with keeping germs out and blood in. So most robot designers feel compelled to use only rotary joints and to call them "elbow," "wrist," etc., despite their many disadvantages in programming, accuracy, rigidity, and size. (Nevertheless, even the Unimate had one linear slide!) Machines designed as machines, by engineers without this compulsion, have both kinds of joints; usually using linear slides for position and rotary joints for orientation.
Cartesian Robots
Gradually good engineering raised its compelling head and more and more CARTESIAN robots were developed, so called after the mathematical co-ordinate system invented by the French mathematician and philosopher Rene Descartes in 1637. (Among his achievements was analytic geometry and the independent invention of calculus, whose better known independent inventor was Isaac Newton.) A convenient way to describe vector motion components in Cartesian machines is north-south, east-west, and up-down for position, and roll, pitch, and yaw (from airplane language) for orientation. Very useful on the telephone. For the mathematically pure one says +- X, +- Y, +- Z, +- alpha, +-beta, +- gamma, and speaks learnedly of coordinate transformations. (To program a jointed arm, anthropomorphic, machine its computer must really perform coordinate transformations.) One of the benefits of Cartesian robots is ease of programming. Since the linear axes are at right angles to each other there is no cross coupling. For point to point work with a small number of positions, limit switches and a very simple computer are all that are needed, typically a PLC. If many positions are needed, only a basic machine tool point to point numerical control system is needed. If continuous path motion is needed, a standard machine tool Computer-Numerical-Control will do. However if "teach mode" is needed for spray painting or the like (see Teach Mode below) there is no programming advantage. A major benefit of Cartesian robots is the ability to make custom combinations of standard motion modules. Another advantage of Cartesian robots is the ability to make a branched configuration having two or more independent motions and grippers. For example, one branch may remove a finished part while the other is picking up a fresh part. A major advantage is the ability to provide very large working space. MOBOT made one robot with over 300 feet of motion along a row of machine tools and made several with 20 feet of vertical motion. At MOBOT Corp. we designed and manufactured such non-anthropomorphic machines and sold them successfully for use in real factory jobs in major corporations, some in the computer industry. We generated credibility partly by de-bunking the hype about anthropomorphic robots.
Robots versus People
From the robot marketers' point of view, the competitors of robots are people. People need little capital investment and can be laid off when orders drop; they are easily programmed; they are easily replaced when in need of maintenance; and they are incredibly dexterous and adaptable compared to any machines. On the other hand, people come in a narrow range of sizes. Between a fragile, small woman and a brawny, large man the strength ratio is less than five to one and the reach distance ratio is less than two to one. Both are equally sensitive to heat, fumes, and noise. Machines, on the other hand, range from micro-manipulators for use under a microscope to steel mill cranes and can be made insensitive to almost any environment. MOBOT Corp. was one of the first to sell Cartesian robots, starting in 1973. As my own robot salesman I found that proposals for tasks easily done by humans were very difficult to sell, but proposals for tasks for which humans required walking, lifting heavy objects, long reach, heavy tools, etc. were much easier to sell. In short, I could sell robots to handle large, heavy loads through long distances. We enlarged our machines and succeeded. The last robot sold before the company's acquisition carried 300 pound jet engine part fixtures and loaded them with extreme precision into a row of large NC lathes spaced along a 400 foot row. Try that with jointed arms.
Pseudo-Robots & Pseudo-Achievements
Pseudo robots were made which were simple machines clothed and painted to look almost human and built to do trivial tasks like delivering a drink and talking via a hidden tape recorder. They were presented with the false implication that they were capable of other tasks as well. Real robots were, and still are, given exhibition tasks to imply capabilities beyond the truth. They have been video-taped "caring for the sick," "making dinner," etc., with all those essential parts of the job which they cannot do, left out. These practices still generate research grants, however. My favorite boondoggle projects are walking robots. I have met a person who gets grants to emulate a horse.
The Bubble
The robot companies' stock prices zoomed. A major brokerage firm had an analyst who, in all sincerity, was devoted to promoting the boom. Robots became high fashion in journalism. In the end, the phoniness served to discredit robots to serious people and the bubble burst. Why? Hype. There was gross exaggeration, by some, of the nature and capability of robots ("electronic brains" etc.) There was the will to believe that a good thing was about to happen. There was, and is, the pervasive, and deliberately planted, confusion of science fiction and science truth. There was an unusually high level of professional conceit among the technical promoters. There was the usual practice of hype, with wishful thinking, to promote money and careers. A grey area exists in which hype give way to hoax; perhaps the difference is the intent of the speaker. Civilization has a long history of bubbles; the book "Extraordinary Popular Delusions And The Madness of Crowds" by Charles Mackay is a classic study.
The Reality
Are robots a passing hoax, then? No. There are real, productive robots which are valuable, cost efficient, smart machines. Some are used as part handlers to load and unload fabricating machines such as machine tools. Some are used as tool handlers, spot welders, arc welders, paint sprayers, de-burrers, and sealant dispensers. Some are used in automatic assembly machines. More and more automatic or remotely controlled machines are called "robots" or "robotic" to benefit their makers and buyers with the glamor of the word "robot." All are intended to replace people at net savings to the manufacturers who invest their cost reduction budgets in robots.
Sensor Controlled Robots
Many robot tasks cannot be programmed exactly in advance because real world operating conditions cannot be predicted exactly. For example, consider the requirements to pick up the top part from a stack of parts, or put down a new part on a stack. Since the part thicknesses vary, one cannot predict the exact position of each part. The same goes for parts in a tray, the location of a conveyor fixture, etc. Means must be engineered to position the robot at the position really required. For another example, consider the problem of putting a part into a very close fitting holder such as a lathe collet, when the clearance between part and holder is less than the positioning accuracy of the robot. Adaptive means other than brute force accuracy are required.
Externally Controlled Robots
Most robots are internally programmed and merely go through the programmed cycle when given a simple start signal. However there are many robots which combine an internal program with obedience to commands from outside. Here are some examples: Automatic storage and retrieval robots (automatic warehouses) are commanded from outside either to transfer an object from a designated pickup point to a designated position in storage or vice versa. The detail motions are internally programmed. Such robots are made in sizes to handle "objects" from tape cassettes to pallets carrying diesel engines. The commands may come from a computer which controls a larger operation, in which case the robot computer and the computer which commands it are said to form a hierarchy. Remotely controlled robots include satellites and space probes which send data by radio back to their human controllers ("telemetry") and receive commands by radio from their human controllers. The detail actions are internally programmed. It is argued that such remote control makes it unnecessary to sent people into space. Mobile robots travel around under some combination of automatic control and remote control. Among them are vehicles for mail delivery, surveillance and police tasks, material transportation in factories, military tasks, and underwater tasks like inspecting pipelines and ship hulls and recovering torpedoes. Although the word, robot, is used at almost every opportunity because of its prestige, it is usually not applied to machine tools. However many machine tools are computer controlled and the control program is replaced for each different product to be made by the machine tool.
Active Homing Guidance
It is possible to add sensors to the robot servos so that the final position of the END EFFECTOR is where the sensors command after a staging position which the program commands. I call such systems "active homing guidance." Typical sensors are limit switches and proximity sensors (optical, acoustic, inductive, capacitive, pneumatic) and in some cases, television.
Passive Homing Guidance
For the ultimate in precision positioning it is possible to make the end effector "permissive" or "compliant" so that forces on the part or its gripper displace moveable portions of the end effector assembly to permit self alignment. In many cases there is no final error whatsoever, the gripper is truly positioned where it should be, plus or minus zero. I call such action "passive homing guidance." In this case, open loop mechanisms out-perform closed loop servos. Heresy! Examples of passive homing guidance are described and illustrated in Chapter 14 of my book, "Designing Cost-Efficient Mechanisms." Homing guidance makes the difference between many successful robots and impossibility.
"Teach Mode"
An early concept in robot development is programming by "teaching" the robot, another anthropomorphic word. The robot is equipped with a switch or other transducer assembly near its end effector. (Sometimes teach switches are separately supported at the end of a flexible cable.) The switch assembly is held by the programming person and moved along the desired motion path while the transducer outputs control the robot motors to follow the transducer assembly. The control computer records the electrical signals generated by the feedback transducers and then regenerates the motion commands. An application of robots where teach mode is indispensable is spray painting. No one can mathematically define the path of a spray nozzle to produce a finish on a refrigerator door without runs or streaks. However a skilled human in grungy overalls and no great power of verbal articulation can move the nozzle to produce perfect results. The solution is to give this human a real nozzle connected to a set of transducers and ask him to paint a real part by hand. Once. The nozzle is then transferred to the robot which reproduces every flick and twist and sweep of the human painter's hand.
The Convertible Robot
One of the hype fantasies was that a robot, being similar to a human, could be easily reassigned from one job to another merely by moving it across the factory floor and plugging in a new program, just as a human is reassigned to a new task and given a new instruction. This conceit is not just an exaggeration, it is flat out not so. Why not? The first reason concerns tooling. You and I have standard tooling called hands and fingers, eyes and touch as transducers, brains as controllers, hundreds of muscles as actuators, and tens of joints as axes. Not only can these tools of ours do remarkable feats of dexterity but they can-and often must- pick up and use extension tools like pliers and scalpels. Robot tools are primitive in comparison and therefore must be made to suit each particular task. (Yes, there are always fatuous R&D efforts to emulate the human hand. Lots of luck from a very experienced robot tool inventor and designer.) Robot controls are primitive in comparison with human controls and robot articulation (axes and actuators) is a tiny fraction of human articulation. The second reason is installation. The machines, conveyors, part magazines, etc. working with the robot must be arrayed in a pattern within which the robot can successfully reach and work. The robot must be accurately positioned within that pattern; it does not have eyes and a brain to adapt to it. Guards and interlocks must be provided to prevent a sudden machine defect from causing mechanical damage or human injury. A master programmer must control both the robot and the associated devices. Often the robot's own programmer is in the bottom layer of a hierarchy of controllers. In short, the robot task itself requires extensive engineering aside from the robot itself. I know of no such multiple purpose robot usage. Robots can be re-programmed easily for variations within task (e.g. spot welding different chassis); but they can not be converted from task to task except as a long term change with a corresponding investment.
Economics
Some robots are cost-justified solely on labor savings, some are cost-justified, at least in part, by the superior uniformity of machine work over human work, and some are cost justified by the savings (and other benefits) in not exposing people to dangerous working conditions. The cost of a robot is approximately the workman's compensation cost of an industrial accident, and there is no assurance that there will not be another accident to the replacement worker. I have read a lot about robots which "improve the quality of life" in factories, but in all my experience I never met a manager with much of a budget to "improve the quality of life" except as a means to improve profits or to conform to law or union pressure. Please excuse the cold blooded but realistic language. The real economic significance of the robot is that the robot itself is a multiple purpose smart machine, produced in quantity, and therefore more cost efficient than a special machine developed to do each kind of job.I'm sorry you will not get a robot to do your housework. Housework is boring and far below your human capability but it is vastly too complex for any smart machine. Automatic, single purpose machines, like dishwashers, you already have and you can count on more and better as time goes by, but a housework robot? No.
Lawrence Kamm
1515 Chatsworth Blvd.
San Diego , California 92107, USA
e-mail:
For further informations in German please click here!