Evolving Robust Robot Team Morphologies for Collective Construction

Abstract

This research falls within evolutionary robotics and the larger taxonomy of cooperative multi-robot systems. A study of comparative methods to adapt the behaviors and morphologies of simulated robot teams that must solve a collective construction task is presented. Multiple versions of an indirect (developmental) encoding method for the artificial evolution of (team) behaviors and morphologies were tested. The indirect encoding method was able to adapt team morphology (number of sensors) and behavior (ANN controller connections and weights) that out-performed a team with fixed morphology and adaptive behavior. Results also indicated that the developmental method was appropriate for evolving controllers that were able to generalize to a range of team morphologies that solved the collective construction task with a high degree of task performance