Evolving co-operation in autonomous robotic systems

There are primarily two ways of producing behaviour based robots capable of achieving complex tasks, either produce a single very sophisticated device or distribute this sophistication across a number of co-operating simpler devices. While the authors adopt the latter approach, both pose the problem of how to design the control architecture once the task becomes complex. This paper discusses automation approaches based on evolutionary techniques. In particular the authors look at evolving a solution to a co-operative object relocation task previously designed manually with a behaviour synthesis architecture. Preliminary results and a discussion on the application of these techniques is given.