Can't Anyone Make a Decent Robot? I'LL EVEN TAKE R2-D2
By Richard A. Shaffer

(FORTUNE Magazine) – I've always loved robots. An impressionable 9-year-old when The Day the Earth Stood Still hit the local movie theater, I practiced saying "Klaatu barada nikto" to keep my cigar-box-and-Erector-set Gort from destroying the planet. In my early teens, inspired by the writings of British psychologist W. Grey Walter, I spent a year of evenings and weekends--plus all of my allowance and paper route earnings--creating what I called a robot turtle. (Others saw it as an upside-down dishpan on wheels.) My automated pet wasn't too smart. A few vacuum tubes, relays, and capacitors served as its nervous system and memory. A creature of wheels, gears, motors, and optics scavenged from surplus World War II bomb sights, my turtle was vaguely like a moth; it could follow a light. When its batteries were low, it could find its way to a charger. Mostly, however, it wandered fecklessly, bumping into the world around it, backing up and starting off in a new direction. But I thought it was marvelous and was convinced that soon, surely before the end of the century, household robots would be commonplace.

Forty years later there's a spate of news that suggests, if only at first glance, that prime time for robots may have arrived. I've read about the rush of buyers for Sony's new robot dog, Aibo. The 3,000 units allotted to the Japanese market sold out within 20 minutes of introduction last month, and an additional 2,000 mechanical canines are now gone too, despite a price of $2,500. Last Christmas, Lego's $200 Mindstorms Robotic Invention System was a huge success. Dozens of models of robots intended for individuals are on the market, from Movit's $24.95 Hyper-Peppy (www.elekit.com) to Hyperbot's $249 Chip (www.hyperbot.com) and the $2,995 CareBot from Gecko Systems (www.geckosystems.com). A hundred robots from dozens of countries are expected in Stockholm next month for the third annual Robocup (www.robocup.org), the World Cup for robots. Friendly Machines of Israel is selling an automated lawn mower, Robomow. Eureka has created a radar-guided vacuum cleaner. And, in what may be a harbinger of a future Furby, Hasbro recently signed a development alliance with IS Robotics of Pawtucket, R.I., whose principals are mobile robotics veterans from MIT.

Sad to say, then, that despite all the news, we're actually not much closer to a good household robot than when I was a boy. The problem that hasn't been solved is the same one that confounded my turtle--for a household robot to know its place, it would be essential for one to know where it is, and that's remarkably difficult.

The usual approach of robot builders has been to cram many cameras, sonar detectors, heat sensors, and accelerometers into the mobile machine and hope that, given enough input and computing power, the robot will develop a sense of place. It's been a reasonable if not entirely satisfactory approach for infinite-budget tasks such as Mars exploration. The household android, however, must make do with less.

Consider my recent experiences with Cye, a small $695 robot I bought as a gift last Christmas and ended up keeping to play with. Bright yellow and low-slung, with two spiky outboard wheels, Cye looks like the foot of an upright vacuum cleaner. (See it online at www.personalrobots.com.) Using the mouse on my computer, which is linked to the robot via radio, I can guide Cye as he rolls around the world and even set up paths he can follow. His chief talent is an ability to find his charger, as my turtle did four decades ago, but Cye is far more intelligent. Instead of merely scanning the horizon for the charger's beacon and homing in, Cye keeps track of his coordinates in a mental map of the surroundings and computes his return path.

At least that's the idea. On a carpeted floor, Cye is proficient. On wood or tile, however, his wheels slip, and he gets disoriented. And, at least for now, Cye is as good as they get.

Digital dog Aibo is full of entertaining tricks, following a ball with the camera in his nose, responding to sounds with microphones in his ears, and standing up again when he falls. But Aibo also walks, which is sure to result in unpredictable motion and keep him from finding his way to the doghouse. He probably doesn't even know where to begin looking because, unlike Cye, Aibo has no system for dead reckoning.

Would a compass help Robo Rover and his kind? Not as much as you might suppose. Digital compasses can be had for as little as $50 each in quantity, but they're accurate only within two degrees, so that over a distance of 50 feet, a robot could be too far out of position to make it through a doorway. Worse, magnetic fields from the robot's motors and circuitboards are likely to be stronger than those from the Earth, compounding the inaccuracy. A global positioning system would be of no use. Portable units are now available for less than $200, but GPS doesn't work indoors and is accurate only to within a yard or so.

Perhaps the answer lies in miniature, inexpensive gyrostabilizers, or in more powerful computers. But a viable realization of my fantasies still lies years off. Linked via radio, the 500 MHz personal computer of today can provide a robot with brainpower somewhere between an ant and a lizard. The IQ of a dog lies ahead of us by a microprocessor generation or two.

RICHARD A. SHAFFER is founder of Technologic Partners, an information company focused on emerging technology. To join his e-mail distribution list, send e-mail to list@technologicp.com. For an expanded version of Watch This Space online, visit www.tpsite. com/tp/fortune/. Except as noted, Shaffer has no financial interest in the companies mentioned. Send comments or questions to shaffer@technologicp.com.