Sign In

Communications of the ACM

ACM TechNews

Robots 'evolve' the Ability to Deceive

View as: Print Mobile App Share:

Prey robot (left) and predator robot (right) coevolve in competition with each other (earlier generation of robots shown in image).

Credit: EPFL Laboratory of Intelligent Systems

Robots programmed to find food concealed their performance from other robots by the 50th generation in a research project at the Ecole Polytechnique Federale de Lausanne in Switzerland.

The researchers equipped the small, wheeled robots with artificial neural networks, and the robots scored points for staying close to food and lost points for being near poison. The robots also could flash a blue light and detect it with their cameras.

"Over the first few generations, robots quickly evolved to successfully locate the food, while emitting light randomly," the researchers say. The team copied and combined the artificial neural networks of the most successful robots to evolve new generations, and made a few random changes to their code that were in line with biological mutations.

After a few hundred generations, most of the robots were no longer drawing attention to themselves when near food by flashing their blue light, and they also evolved to become either attracted to or repelled by the light. The results of the project could help provide more answers regarding the development of biological communication systems.

From Technology Review
View Full Article


Abstracts Copyright © 2009 Information Inc., Bethesda, Maryland, USA


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account