An experiment in mixing evolving and preprogrammed robots

Abstract

Artificial evolution of robot behavior is commonly conducted in environments containing a single robot or multiple robots that are all controlled by evolving behavioral logic. In this paper, we take a novel approach and study how the presence of preprogrammed robots affects the evolutionary process and the solutions evolved. We evolve behavioral control that enables robots to forage. The robots are situated in an environment that contains a nest and a number of prey. The robots must either push or carry the prey to the nest. We analyze the behaviors evolved in mixed setups in which one or more preprogrammed robots are present. We compare these behaviors to behaviors evolved in non-mixed setup in which no preprogrammed robots are present. The results show that although the evolved robots do not use their capacity to communicate, they do collaborate with the preprogrammed robots. We find that the performance of some of the solutions evolved in the mixed setup is higher than the performance of homogeneous groups of robots.