Seeking Wider Acceptance of Robots, Designers Create Toys
Use of artificial intelligence in robots key to developing consumer and military markets.
SOMERVILLE, MASS. — Imagine playing with a "smart" ball that cowers and cries with fear when it rolls into a dark area. Or a baby doll that looks at you and gurgles happily when you rock it, and burps after it drinks from a bottle.
These toys and others still are a couple of years down the road. But they represent a wave of new toys with the ability to interact with the person playing with them. They also represent the new breed of robots.
They're not clunky robots like "Rosie" of the Jetsons cartoon show or R2D2 of the "Star Wars" movies. The robot baby looks and acts uncannily like a real child.
"We feel like we made a baby rather than an interactive baby doll," said Colin Angle, chief executive officer and co-founder of IS Robotics Inc. (ISR) in Somerville, Mass., which created the doll and ball. "The difference is that you can turn this baby off when it cries."
The baby doll has animatronic areas in its face that allow its mouth to open, its cheeks to draw tight when it cries, and its eyes to squint. With sensors, the baby doll can understand 12 different actions including patty-cake, a hug, a bottle in its mouth, a tickle, and being turned upside down. It can react by looking at its playmate, crying, laughing, and sucking loudly on a bottle. The doll runs on AA batteries.
The toys still are experimental. Mr. Angle says a major US toy manufacturer is looking into commercializing them some time in 2000 or 2001. The balls could have different attitudes or personalities, and would cost about $20. A fully featured doll might sell for $250, but a doll with fewer features could sell for as little as $50.
Angle and his colleagues used artificial-intelligence technology devised at the Massachusetts Institute of Technology's Artificial Intelligence Lab to make toys that know, intuitively, how people should play with them. The toys then throw the player curves to make play more interesting. For example, the ball can round a corner or hide under a table.
"It's a ball with an attitude," says Angle, an MIT graduate.
He was inspired by the run-away success of Tamagotchi, the virtual pet on a key chain that its owner must care for, or it "dies." "When I saw the success of that, I built the ball," Angle says.
Angle sees toys as a way to popularize robots, which generally are considered hulking, awkward gizmos for industry. To date, most of ISR's business has been with US government contractors. The company makes experimental robots to find and help disarm land mines. Those can work in various harsh environments, including under water and on a sandy beach. It also is making a robot for an oil company that will help maintain oil wells. That robot will be able to live in the underground tubes where the oil travels to the surface of the earth.
"Robots have been successful in factory lines, where precise movement of a robot arm has a lot of value," Angle says. "But robots are not widely accepted outside that area."
"For robots to really make inroads into affecting our day-to-day lives, a robot must first be able to operate in our world, with its steps and curves," he adds. "And the power supply must run long enough before requiring recharging."
Robots already are creeping into our homes. Even everyday appliances like breadmakers can be considered robots. Angle says any device that perceives its world and takes action based on that perception is a robot. He figures toys will be a key to unlocking wider use of robots in homes.
"The challenge is what to do with interactivity so a player is engaged by the toy, so the imagination is stimulated," he says.
ISR's technology differs from its competitors in how it tells the robots to act. It breaks down behaviors into components. For example, to walk through a door requires standing, balancing, walking, avoiding obstacles, recognizing the door, and getting through the door. The company created a powerful software program to describe common movements with about 4,000 behaviors that are run on one microprocessor.
When the company started in 1990, the dominant technique for moving robots was to make a model of the environment using the senses, and then apply complex algorithmic formulas to issue commands to joints in the robot's body.
"This is hard to do, and the model of the environment is usually wrong," Angle says. He added that ISR's technique is used in most robots nowadays.
To take the technique beyond expensive research and factory robots, ISR scaled down its software a bit and used less expensive electronics and sensors.
ISRs experimental toys go beyond Tickle Me Elmo and electronic Barney dolls. With an added new box from Microsoft Corp. between an electronic Barney and a TV set playing the Barney show, the doll can sing along with the TV tunes.
"What Microsoft did was to create a robot that interacts with the TV," Angle says. "But I'm making a toy that can interact with me."
He hopes the friendly toys will mute the science-fiction notion of robots going awry in the home. A helpful robot could turn on lights, help prepare dinner, and clean up afterward. Says Angle, "There is a market for a robot vacuum that can pick up crumbs after I eat potato chips."