INVASION OF THE SERVICE ROBOTS Like human workers, robots are moving into services -- especially jobs people find dangerous or boring. Their boosters see a much bigger market than in manufacturing.
By Gene Bylinsky REPORTER ASSOCIATE Alicia Hills Moore

(FORTUNE Magazine) – FOR A LONG TIME robots have been stuck on the factory floor, toiling away at such repetitious, brute-force chores as welding car bodies and lifting heavy steel bars. Now they're breaking loose. Like their human colleagues, they are moving increasingly from manufacturing into services. For several years service robots have been at work in nuclear plants, where people risk exposure to radiation, and under the sea, where human divers require cumbersome and costly life-support systems. Today the protean machines are embarking on a multitude of new activities: taking care of the handicapped and elderly, picking oranges, cleaning office buildings and hotel rooms, guarding commercial buildings, even helping cops and brain surgeons. Doctors at California's Long Beach Memorial Hospital have performed more than 20 delicate brain operations with the help of a robot arm that drills into the skull with great precision. In Dallas a year and a half ago, police used a robot to bluff a suspect barricaded in an apartment into surrendering. When the robot broke a window with its scary mechanical arm, the man came running out the front door shouting, ''What the hell was that?'' In U.S. laboratories, more than 1,200 robots perform such intricate tasks as weighing, measuring, and mixing minute quantities of chemicals, medications, and even DNA. The Navy is starting to deploy undersea robots with lobsterlike claws that snip the mooring cables of stationary mines, such as those currently threatening ships in the Persian Gulf. Other fields wide open for service robots include building and maintaining offshore oil rigs, working on construction sites, tending hospital patients, assembling space stations, pumping gasoline, preparing fast food, fighting fires, and inspecting high- tension electric wires. ''We're seeing the birth of a big, new industry,'' says Joseph F. Engelberger, 62, the father of industrial robotics and the principal driving force behind service robots. ''Now that robots can be mobile and are starting to be able to see and feel, service jobs will eclipse the entire manufacturing scene for robotics.'' He envisions service robots soaring to $2 billion in annual sales in the U.S. by 1995, up from about $120 million today, mostly in undersea applications. (Like many such estimates in high technology, that $2- billion figure may well be optimistic.) By contrast, the U.S. market for industrial robots is expected to reach $370 million this year and $1 billion in the mid-1990s. As with almost any new technology, the major innovators aren't the big, established companies -- in this case industrial-robot makers GM Fanuc Robotics, IBM, and Cincinnati Milacron -- but small companies such as Engelberger's Transitions Research Corp. (TRC) of Bethel, Connecticut. (About the only exception is Westinghouse, which provides service robots for nuclear utilities.) In the late 1950s Engelberger ignited the industrial robotics revolution by starting Unimation Inc., the first robot company. TRC, a three-year-old privately held company with annual sales of about $1.5 million, is Engelberger's entry in the service-robot sweepstakes. The last time around, the Japanese in effect stole Engelberger's baby, the industrial robot. Because he could not afford to patent his robot in Japan, they ran away with it by copying and mass-producing the machines. Japanese robot makers such as Fanuc and Matsushita now control the world industrial-robot market. To keep the nascent service-robot industry from slipping into Japanese hands too, Engelberger has assembled an impressive lineup of big corporate backers for his projects, including Du Pont, 3M, Johnson Wax, Electrolux AB, Maytag, and Emhart, a Connecticut conglomerate that had $2 billion in sales last year. While Japan isn't exactly asleep, as of now it is behind in the underlying technologies. A VISITOR TO TRC can be greeted by a squat three-foot-tall robot, a test- bed for an automatic vacuum cleaner, rolling by Engelberger's office with its electronic innards exposed. Research on the vacuum cleaner is sponsored by Electrolux, the big Swedish consumer electronics concern. Designed to clean malls, supermarkets, factories, and airports without human assistance, it will sell for about $20,000 when it goes on the market next year. According to TRC, . outside surveys show that a business with 35,000 square feet -- the size of a small shopping mall -- to clean could use the machine economically. Other Engelberger robots include HelpMate, a nurse's aide that will deliver meal trays to bedridden hospital patients. Along the way, the wheeled robot will take elevators and negotiate hallways by itself, doubtless startling unwary visitors. HelpMate will be tested starting next January at the Danbury, Connecticut, hospital; Engelberger plans to put it into mass production in 1989. It will sell for about $25,000 and will pay for itself in 2 1/2 years if it is used 24 hours a day, according to TRC-sponsored studies. Engelberger's ''ultimate robot,'' as he calls it, will be a $50,000 household helper -- assuming he can raise about $20 million to develop it. He sees this robot as a high-tech butler that would prepare meals, clean the house, cut the grass, clear the driveway of snow, and even fix household appliances such as refrigerators and washing machines. ''You'll get value and return on investment compared with hiring a practical nurse at $9 an hour to help your old mother get around the house, as I do now, for example,'' Engelberger says. One snag: He hasn't yet figured out how to teach it to make the beds. Not everyone agrees that an all-purpose household robot will work. David Nitzan, director of the robotics lab at SRI International, thinks that it would make more sense technically to concentrate on single applications -- a bathroom-cleaning robot, for example. But no one disputes that service robots are expanding. Among other things, Nitzan and his group are working on the vision system for a robot that would automatically sort packages for the U.S. Postal Service. Why the service robot surge? For one thing, progress in core technologies -- vision, mobility, controls -- has been rapid. ''It's a golden time,'' says William Whittaker, a senior scientist at the Carnegie Mellon University Robotics Institute in Pittsburgh, a leading developer of service robots. ''A couple of years ago my feeling was that there was very little insight into either what to do or how to do it. But now those contributing or enabling technologies are there for the most part. There's the magnitude of the market, the potential and the inevitability of the technologies. It's as foregone as computing was 20 years ago. It has the same feel to it.'' There are demographic pressures as well. Many service-robot suppliers see the rapidly aging U.S. population as a huge, new market. People over 85 now make up the fastest-growing segment of the elderly in the U.S., and 200,000 older Americans a year break their hips. Says Karen G. Engelhardt, director of the Health and Human Sevices Robotics Laboratory at Carnegie Mellon: ''Never before have we seen a technology with such promise and potential to help this large and growing population in ways they never could be helped before.'' BECAUSE SERVICE ROBOTS are often able to make repairs under hazardous conditions faster than people can, for example, they may deliver the increased productivity that has generally eluded the service sector. Shortages of service workers such as nurses are another force driving the advent of service robots. So are corporate policies and regulatory pressures against placing workers in dangerous settings. Since 1971, for instance, 54 deep-sea divers have been lost in North Sea offshore oil and gas operations. The shift toward robot submersibles has turned a lot of divers into shipboard robot operators -- a much safer occupation. In nuclear plants, robots toil for hours at a time in highly radioactive areas in place of hundreds of employees, called jumpers or glowboys, who worked in short relays so as to minimize their exposure. In space, robots can significantly speed assembly of big structures like NASA's projected space station for the 1990s. In those three settings -- in nuclear plants, under the sea, and in space -- cost savings are a big factor as well. Supporting deep-sea divers and astronauts working outside their spacecraft can cost up to $100,000 an hour. When a nuclear plant shuts down, replacing the lost electricity can cost a utility an estimated $500,000 a day; robots have already helped shorten those shutdowns. The military, whose primary concern isn't economics, sees in the new robots a way to maintain high-tech superiority over more numerous enemy armies on a battlefield. The Pentagon is spending hundreds of millions of dollars a year to develop small unmanned tanks, intelligent robotic undersea vehicles, and flying spy robots that fill the gap between airplanes and satellites. Robert Finkelstein, president of Robotic Technology Inc. of Potomac, Maryland, who works closely with the Pentagon, says research and development outlays for robot systems will soar to $3 billion a year by 1995. Both robot builders and users now recognize that the new robots can do a ! lot of things humans can't. Anthropomorphism -- having robots imitate human activities -- has fascinated robot buffs ever since the Czech playwright Karel Capek introduced the word ''robot'' into the lexicon. In his popular 1921 play R.U.R., for Rossum's Universal Robots, Capek made robots look like mechanical counterparts of man. He derived the word from the Czech robota, which means work -- including forced labor. Anthropomorphism is appealing, but many robot builders think it is usually a wrong track in robotics because it fails to take advantage of some remarkable man-made devices and technologies. Although highly dextrous, the human arm and wrist, for instance, cannot twist completely around. A robot hand can. ''We can try all we want to imitate mammals,'' says TRC's Engelberger, ''but no animal has a built-in ball bearing.'' A robot arm that rotates in all possible directions, for example, is more useful than an imitation human arm that is limited by its joints. In a remarkable demonstration of robotic agility, a nuclear-plant robot built by Odetics Inc. of Anaheim, California, not only walks on six legs but also can extend itself from a height of five feet to a total of 14 feet, becoming a kind of a mechanical giraffe. It can hang from ledges inside nuclear plants, working upside down. By automatically changing the gripper in its hand it can do a variety of tasks, from changing light bulbs -- a huge job in a nuclear plant -- to lifting objects that weigh up to 300 pounds. Similarly, surveillance robots can do things a human guard cannot. Thanks to microwave vision, they can see through nonmetallic walls and in the dark can spot an intruder as far as 130 feet away. Most robots in use today in space, under the sea, and in nuclear plants are operated from a distance by human workers as extensions of themselves. More often than not, the robots are tethered by a cable to the control station; the cable transmits electric power and serves as a communication link to the robot. Robot builders call this telerobotics, or teleoperation -- an extension of human sensing and manipulating capability. WHILE TELEOPERATION obviously keeps people out of hazardous environments, it has drawbacks. Operators can have trouble controlling remote robots and monitoring exactly what they're doing -- for instance, when they're working undersea in murky waters. For greater versatility, a robot should work on its own, with minimal human supervision. Says Robotic Technology's Finkelstein: ''The essence now is intelligent control systems. The technology for teleoperations is here now. The technology for automatic operations is being developed. If you have a tank as smart as an ant that knows how to tell an enemy from a friend, that could revolutionize warfare.'' Currently, the U.S. Army is developing a walking truck to traverse roadless terrain, a contraption with fat, bent legs and a wriggling body that make it look like an immense, unearthly insect. The growing military applications of robots promise civilian fallout that could pay off for companies big and small. For example, military work on robots that navigate by themselves could be applied to trucks, automobiles, and factory vehicles. Early designs for NASA's space station assembly robot, called Flight Telerobotic Servicer (FTS), resemble a refrigerator with a flat head and four arms. FTS will be a hybrid between a telerobot and a fully autonomous machine. An astronaut can run it in two ways. In the teleoperated mode, the physical motions of an astronaut's hand will be reproduced by the robot's mechanical arms, which multiply the astronaut's force. The astronaut will work in shirtsleeves inside the space station; the robot will be outside. Alternatively, the robot could operate ''teleautonomously.'' The astronaut would transmit complex commands for it to interpret and execute. The astronaut, for instance, could command the robot to repair a nearby spacecraft. The robot would dock with the satellite and do its job without any further commands from its operator, who would watch the robot on his TV screen and intervene if needed. Six NASA contractors -- Westinghouse, Grumman, Fairchild, Lockheed, Martin Marietta, and United Technologies -- have just submitted preliminary design plans for the FTS robot. Giulio Varsi, manager of automation and robotics for NASA's space station office, talks enthusiastically about developing space robots that ''learn as they go along'' -- intelligent automatons with sufficient vision and a fine sense of touch that can modify their actions as circumstances change. Such technologies would help improve earth-based service robots. The initial FTS robot will cost about $200 million, in part because of its highly complex software, but subsequent versions could probably be produced for about $10 million apiece, Varsi says. THE MOST SURPRISING demonstration of robot autonomy so far will take place as early as October at New Hampshire's Lake Winnipesaukee. In a game of wits, government and University of New Hampshire researchers will pit two untethered undersea robot craft working together against men in boats, who will try to detect the robots with sonar and other instruments. For their part the robots will work in tandem, using microprocessors guided by artificial intelligence software. The robot craft will attempt to elude their human pursuers by sacrificing one robot and having the other escape with TV pictures and other data it has gathered. The key to making robots smarter lies in computer power that can enhance their intelligence, which is so far rather dim. Carnegie Mellon's Whittaker says that some of the service robots his institute has built have only reached the ''intelligence of a worm.'' Robots as brainy as the pair in Star Wars, the timid android C3PO and its barrel-shaped electronic sidekick, R2D2, are at least 50 years away.

One reason: Progress toward giving robots humanlike sight and hands has been excruciatingly slow. In the case of sight, digital computers have trouble recognizing patterns that people spot almost intuitively. After about a decade of work, researchers at Stanford University who are trying to impart vision to robots have just scored what they consider an impressive victory. Their vision system can recognize about two dozen different objects -- airplanes of varying shapes, for example -- as belonging to a single class. A University of Utah- MIT project in progress for seven years has yet to produce a fully workable hand, although researchers have constructed a computerized prototype with tendonlike wires and four fingers run by 32 miniature motors. Both research teams have come to appreciate the awesome complexity of human senses and information processing. Says Stanford researcher Thomas O. Binford: ''The retina of one eye has roughly 100 million specialized vision cells and four layers of neurons, all capable of doing about ten billion calculations a second.'' All told, about 60% of the brain's cortex, the so- called thinking cap on top of the brain, is involved in handling visual information -- a computational task that it would take 100,000 Cray supercomputers to handle, Binford estimates. If researchers can get robots to see clearly and to understand where they are -- another classic problem -- that would make possible seeing-eye robots, among other things. The elusiveness of perfection does not deter the practical types, however. ''University researchers sometimes unnecessarily complicate ^ things,'' says Engelberger. ''We take bits and pieces of available technology -- ultrasound, infrared, TV cameras -- and we tell our robots beforehand where to go.'' ENGELBERGER PLANS to get around the ''Where am I?'' problem in his cleaning robot by using what he calls the Hansel and Gretel concept. In a typical operation, the robot begins by circling the perimeter of a room, bouncing sonar off the walls to locate itself. As the robot goes along, the Hansel and Gretel scheme comes into play. Those fairy tale tots marked their path in the forest by leaving a trail of bread crumbs to guide them back out. Birds ate the crumbs, of course, or there wouldn't have been a story. Each time Engelberger's cleaner makes a circuit, it will drop tiny pieces of fluorescent paper -- ''bread crumb equivalents,'' Engelberger calls them -- to one side, creating a parallel, inner circle that it will follow the next time around. It's much like mowing a lawn in a continuous spiral from the outside in. As the robot follows the path, its scrubbing brushes will clean up the old trail just as Hansel and Gretel's birds did. This or similar approaches are a lot easier than trying to build up a comprehensive picture of the world inside a robot's brain. As for hands: Many of today's robots can automatically change their ''end effectors,'' tools that take the place of fingers to perform different jobs. Odetics, the Anaheim company that builds the walking robot, has just developed a hand that can grip a wider variety of objects and shapes -- from a pencil to a railroad tie -- than other models. It has two thumbs and one finger; the thumbs can rotate to grasp a payload, just as human thumbs can. Unlike a human hand or the University of Utah-MIT design, the Odetics hand requires no ''tendons'' in the arm or wrist to provide the power to move. AT CARNEGIE MELLON'S Robotics Institute, Bill Whittaker sees the next important advances in service robots coming not in the basic technologies that make the robots possible but in combining them into working robots. Putting that view into practice, his institute has been turning out an impressive flock of robots. One has been used to decontaminate radioactive parts of the disabled Three Mile Island reactor. Another, a hulking brute of a machine about the size of a compact car, has four wheels and four stiff legs to walk with if its steering or the motors that drive its wheels fail. It can work with a hook, a crowbar, a shovel, a saw, a water-jet cutter, and other tools it carries. Stanford University robot researchers have helped bring to fruition an impressive robot system that takes care of at least 13 needs of quadriplegics, from preparing canned soup and serving it to brushing their teeth afterward. During a recent demonstration of its remarkable skills at the Palo Alto Veterans Administration Medical Center, the robot greeted visitors with a brisk ''Hi, Earthlings!'' and proceeded to put on an impressive show. Because the research has been financed with federal funds, the system is available to any company that wants to build and sell it. Stanford and VA researchers figure it can be built for about $50,000. What's coming will be even more surprising: smarter and even more mobile service robots that will be directed by a few simple words or even the gesture of a hand. Says the ever visionary Engelberger: ''The list of service applications will grow because some of the more stultifying, demeaning, and downright dangerous human activities are in the service jobs. Robotizing those jobs is both possible and economically justifiable. This is not just an extrapolation of industrial robotics but literally a new slave class -- mobile, sensate, service robots.'' Ready when you are, R2D2.