Roaming robots

Why Files:

robot vehicles

robot chess machines

big brother

This little four-wheeler (above right) sends signals from its video camera to the hand-held controller.

Wearing official Wisconsin headgear, Lobotomous automatically served hors d'oeuvres. Out on a stroll, Dan Stormont directed Lobotomous with a keyboard.

POSTED 6 AUG 1998. A bread-box-sized robot prowled the floor on "robot night" at the 1998 conference of the American Association for Artificial Intelligence in Madison, Wis. if you lived here you'd be dead nowUnlike most of the student-built robots at the robot confab, this little critter had no human minder.

Many of those minders -- slightly frantic when things went wrong, wearing an idiot grin when they went smoothly -- were directing their robots via attached keyboards. (During competition, of course, the robots had to operate autonomously -- using the logic of artificial intelligence.)

Not this little four-wheeler, a Real World Interface model called the micro-ATRV. Red, purposeful and moving zestfully across the floor, it skirted obstacles, seeming to know exactly where it was and where it wanted to be.

No techno-slave here
I figured some programmer had spent hours coding in the location of every booth, post and door in the exhibit hall. Only later, talking with Robert Pack, who directs software research for Real World Interface did I learn the secret -- this R2D2 wannabe was feeding a video signal to a screen on the hand-held control panel.

It turned out that the intelligence planning those brainy moves was inside a human skull.

It didn't take The Why Files 10 nanoseconds to pose the obvious question: "Can we try?"

With a box controlling speed and direction in our grubby mitts, The Why Files tried to get the hang of guiding the robot by looking only at the video screen. As soon as we got halfway proficient, the battery power ebbed, and the video screen began looking more like a migraine headache than a robot's-eye-view of the convention floor.

Blind, the ain't-I-cute robot turned into a unguided missile, and observing our futile attempt to recover, the robot's owner strode over and hauled it bodily back to base for a recharge. (Had the 'bot been a dog, it would have had its tail you-know-where.)

Aside from the need for better batteries, what did we learn about robots?

Robots are cheap enough for undergraduates to play with.

One form of sensation is seldom enough -- many robots are studded with video cameras, and infrared and sonar detectors.

Navigation is the biggest single problem. It takes a heap of artificial intelligence to tell a robot how to respond when, say, a big foot appears in a detector. (Many of the student-built models were directed to freeze in those conditions.)

Robots need artificial smarts. Making a robot move and detect signals from its environment is much easier than helping its evaluate those signals and devise a course of action. Merely distinguishing a person from a post is quite a challenge for the average robot.

Want fries with  that?Naive navigation
Although robots like complete information, Shar Whisehunt, an undergraduate from the University of Texas, said she was trying to get robots to operate without it. "My focus is dealing with uncertainty. You know about where you parked your car, but not exactly where, and that's good enough." Not so for robots.

I'd rather have a robot in front of me than a frontal lobotomySafety and human interactions are also major areas of investigation, said Dan Stormont, a recent graduate of the University of New Mexico who's now assistant professor of aerospace studies at Utah State University in Logan. "How do you get people to trust a robot?" Stormont asked as he piloted (via computer keyboard) the "Lobotomous" robot back from a tour of robot-served hors d'oeuvres.

"You've got to prove to people that it won't kill them." That's easy with lab-built models at a robot fair -- when confused or confronted, they simply hunker down and wait for instructions. But with robot cars, "You can't just drop dead on the highway," he notes. (The Why Files covered the cross-country travels of one robot car.)

Lone star triumph
The Why Files next talked with undergraduates from the University of Texas. Fresh from their victory in the "Life on Mars" competition, they crowed that no other robot had located the objects that simulated martian life, and placed them in the correct door of the "martian lander."

All this is fine on Mars. But as robots get smarter, won't they take jobs away from humans? Perhaps, but true that robots -- in the form of numerically controlled machine tools -- have been working for 20 years in factories, with no obvious damage to employment. According to Pack, robots are now suited to jobs like cleaning, surveillance and security. "Our goal is not to take any jobs away that anyone likes. We want to take away the jobs that people hate."


-- David Tenenbaum The Why Files

Credits | Feedback | Search