CHICAGO To build a better robot, scientists increasingly are looking to nature, making robots that move and interact socially with cockroaches, slither like a salamander and even learn and make decisions like humans.
The new designs, reported on Thursday in the journal Science, suggest that robot science is finally catching up with science fiction.
One team of European researchers led by Jose Halloy of the University Libre de Bruxelles in Belgium, created cockroach-sized robots that interacted with their living counterparts.
While they did not look like cockroaches, they were coated with chemicals to make them smell like cockroaches.
And they behaved like robots, so much so that they influenced their roach clan into making a bad choice of shelter, choosing a light shelter rather than a darker one.
"What is new here is that the robot is autonomous, it is not remote-controlled by humans, and it acts at the social level in a group living insects," Halloy said in an e-mail.
By changing some parameters, the robotic cockroaches infiltrated the group and influenced its behavior.
"We see them as a tool to explore decision-making mechanisms in group-living animals," Halloy said.
ROBOTS WITH BRAINS
The study is part of a bigger effort to create robots that can respond flexibly to changes in the environment.
Gerald Edelman of the Neurosciences Institute in San Diego, California, has spent the past 20 years working on brain-based devices that simulate the activity of the human brain.
"Of course, the number of neurons and synapses is much smaller. But if you do the thing right, it actually does function for sophisticated purposes, such as perception and episodic memory," Edelman said in a telephone interview.
Because the brain learns by interacting with the environment, building a robot that can learn must incorporate this process. "It's not a brain in a vat. It's the interaction of the body and the environment that counts," he said.
Edelman, writing in Science, describes Darwin VII, a brain-based robot fitted with a video camera for vision, microphones for hearing and a gripper device that can sense how well different steel blocks conduct electricity, a process that functions as a sense of taste.
Darwin VII could learn which blocks had a good taste -- they offered a type of reward -- and which had a bad taste, in an experiment that replicated a value system.
A subsequent version, Darwin X, learned how to find a hidden platform, replicating a common rat study in which they must find a submerged platform in a maze of milky fluid.
In this case, Darwin's infrared detector sent a reward signal each time it passed over the platform. After training, Darwin could go to the platform from any point in the maze.
Edelman said this process can be studied to understand how the brain learns.
In a review of biologically inspired devices, Rolf Pfeifer of the University of Zurich describes robot insects, spiders, snakes, lobsters, dogs, monkeys and humans.
Such robots, Pfeifer suggests, offer the opportunity to study different patterns of movement, locomotion, navigation, orientation, manipulation, imitation and cooperation.
Ultimately, he wrote, these might result in machines that are "adaptable, robust, versatile and agile."
"Exciting times are ahead of us," he wrote.
(Editing by Maggie Fox and Sandra Maler)