Researchers Teach Robots Deception by Mimicking Squirrels

Squirrels aren't exactly the first thing that come to mind when you think of withholding the truth, but researchers have used the bushy-tailed creatures to teach robots to do exactly that.

Researchers at the Georgia Institute of Technology at the behest of the Office of Naval Research have developed technology that can allow robots to deceive each other, much in the same way squirrels do when trying to protect their hidden acorns. The military could one day use the robots to protect ammunition or supply stores in combat situations or on the battlefield.

"We have developed algorithms that allow a robot to determine whether it should deceive a human or other intelligent machine and we have designed techniques that help the robot select the best deceptive strategy to reduce its chance of being discovered," said Ronald Arkin, a Regents professor in the Georgia Tech School of Interactive Computing, in a press release.

Researchers used the behavior of squirrels when storing food to develop the algorithms. Squirrels store their acorns in specific locations and routinely keep an eye on those spots to ensure the food is still there. When other squirrels are near a food store, the possessive squirrel will try to lure them away by visiting spots where no acorns are stored and treating those spots in a similar manner. In nature, this behavior works.

Researchers also used a type of bird found in the Middle East called the Arabian babbler to develop the algorithms. These birds distract a predator bird by joining with other birds and attacking a predator in a fake show of force. The deception technique can inspire the predator to give up the hunt.

To prove that the deception technology works in robots, Georgia Tech researchers dispatched two autonomous robots to play 20 games of hide-and-seek. Researchers lined up colored markers along three potential pathways to places where one of the robots could hide, while the other robot sought to find it. On the way to the spot, the hiding robot would knock down the colored markers. To deceive the seeking robot, the hiding robot would proceed to a hiding spot and knock down the markers, but then change course and hide in another location that couldn't be discerned from the displaced markers. The seeking robot used the presence or absence of markers to find the hiding spot.

The hiding robot was successful at deceiving the other robot in 75 percent of tests, researchers said. Failures occurred when the hiding robot did not knock over the appropriate markers to accurately deceive the seeking robot. "The experimental results weren't perfect, but they demonstrated the learning and use of deception signals by real robots in a noisy environment," said Georgia Tech Institute research engineer Alan Wagner in the press release.

Robotics researchers already have been using the movements of animals, birds, and insects in their design of robot movements for some time, but now animal behavior is becoming increasingly more relevant for the design of robot intelligence. In other work, researchers in the UK are working on computer models of a honey bee's brain to create a robot that can think like the flying insect.

Related posts:

Comments (0)

Please log in or to post comments.
  • Oldest First
  • Newest First
Loading Comments...