Evolving adaptive group behavior in a multi-robot system

Abstract

The field of multi-robot systems is sometimes called swarm robotics when the systems consist of many simple autonomous robots. However, each robot is usually assumed to have no learning mechanism for adapting to an embedded changing environment. Therefore, collective behavior is expected to emerge in the system only through interactions among the robot. This implies that they cannot be coordinating as a group. In this study, an evolutionary robotics approach is applied empirically to a multi-robot system to realize autonomous task allocation behavior as a kind of intelligent swarm robotics. Although artificial evolution has proven to be a promising approach to coordinate the controller of an autonomous robot, its effectiveness in developing beneficial collective behavior in a multi-robot system has not been verified. Several computer simulations are conducted to examine how artificial evolution contributes to autonomous task allocation in a multi-robot system.