November 17, 2006 10:26 AM PST
Researchers unveil a self-aware robot
Scientists said Thursday they created a brainy, four-legged robot resembling a starfish that can sense damage to its body and, on its own, think up a way to recover.
Researchers Hod Lipson and Victor Zykov of Cornell University and Josh Bongard of the University of Vermont made a robot that observed its own motion using built-in sensors in its joints and then generated its own concept of itself, or at least its physical structure, in its internal computer.
It used this internal model of itself to figure out how to walk on its four legs and eight motorized joints.
"In the beginning, the robot starts off and does not know what it looks like. You look at it, and you see that it's a four-legged machine. But the robot itself doesn't know that. All it knows is that it could be a snake, it could be a tree, it could have six legs," Lipson said in an interview.
Lipson said the robot used various movements of its joints, first to generate hypotheses and then to formulate an accurate conception of itself.
The researchers then tested the robot's ability to adapt to new situations--in this case injury--by shortening one of its legs. "The robot knows something's wrong," Lipson said.
Animals can compensate for injury by changing movements, like limping to favor an injured leg. Machines can be programmed to react to a problem in a certain way. But when they are damaged in unexpected ways, they usually are doomed.
This plucky robot responded by generating on its own a new concept of its structure, accurately sensing it had been altered, and then devising a new way to walk using a different gait to compensate for the injury.
The robot's smarts, awareness of itself and ability to adapt on its own separates it from its mechanical brethren.
The study was published in the journal Science.
"We don't really think this is self-consciousness, which is a robot thinking about itself thinking," Lipson said. "But I do think it is moving in the direction of consciousness, like a cat, that kind of level."
Aside from contributing to a philosophical debate, the research has practical implications--giving hope to people who envision sending robots that can adapt to unforeseen circumstances to explore other worlds or the ocean floor.
"There is a need for planetary robotic rovers to be able to fix things on their own," Bongard said in a statement. "Robots on other planets must be able to continue their mission without human intervention in the event they are damaged and cannot communicate their problem back to Earth."
Christoph Adami of the Keck Graduate Institute of Applied Life Sciences in Claremont, California, wrote a commentary accompanying the research titled, "What Do Robots Dream Of?"--an allusion to science fiction writer Philip Dick's 1968 novel, "Do Androids Dream of Electric Sheep?"
Adami described how a robot like this one might perform in unknown territory, exploring the landscape and then "dreaming" of new methods to overcome obstacles it had encountered.
"And even though the robots...seem to prefer to dream about themselves rather than electric sheep, they just may have unwittingly helped us understand what dreams are for," he said.
Reminded of malicious robots and computers turning on their human masters in movies like "The Terminator" and "2001: A Space Odyssey," Lipson was not worried.
"We just pull the plug out of the robot. That's all," Lipson said. "There are more immediate things to worry about than to worry about that."