Evolution Of Communication Gleaned From Robots

Identical animal populations living in identical habitats may evolve different means of communication due to random genetic drift, researchers find. Laurent Keller and colleagues used robot simulations to model complex evolutionary change. The researchers began with 20 populations of identical robots, each equipped with two wheels, a camera, a food detection sensor, a simple information processing program, and a ring that could emit a blue or green light. Each population of 20 robots was placed in an arena containing a food source. 

The researchers ranked each robot according to the amount of time spent on the food source, and used a common statistical method to select 100 robots' programs or "genes" for reproduction. Because the "genes"--which encoded specifications of the robots' neural controllers, responsible for processing sensory information and producing motor actions--were initially set to random values, the robots behaved unpredictably at first. But after 1000 generations, all 20 populations emitted light to indicate food location. In approximately half the populations, the robots emitted a signal only in the presence of food, while the other population also emitted a different color light in areas without food. 

The one-signal populations found food faster on average than did the two-signal populations; however, one-signal robots fared the worst in competitions between robots with different strategies. Additional tests revealed that signaling differences occurred early in the robots' evolution. Evolutionary divergence in communication is commonly attributed to habitat or mating differences. The results suggest that such evolutionary pressures are not necessarily required for animals to evolve different means of signaling, according to the authors.

Contacts and sources:

PNAS News Office 
Proceedings of the National Academy of Sciences

Article #11-04267: "Historical contingency affects signaling strategies and competitive abilities in evolving populations of simulated robots," by Steffen Wischmann, Dario Floreano, and Laurent Keller