Interleaving simulated and physical environments improves evolution of robot control structures

Abstract

Control structures for physical robots can be evolved in simulated and physical environments. In this study, the interleaving of simulated and physical environments during the course of the evolution of a control structure was examined. This method was compared to the method of 'fine tuning' in a physical environment a control structure that has involved in complete simulation. Interleaving physical and simulated environments improves performance of the eventual control structure. Possibly, this method allows for the evolved control structure to incorporate and retain advantageous behavioral patterns from both environments.