Co-evolution of Morphology, Control and Behavior Master project

Currently, the maneuverability of robots is obtained by human design and is specific to each robot. There are no generic ways of designing robots with maneuverability skills. In this paper we investigate the use of co-evolution of body and mind in virtual robots to optimize maneuverability. Robots were successfully evolved to respond to steering commands and speed commands. It has been demonstrated that specific morphologies and controls are required to suite the maneuverability criteria. The robots developed for this project are designed to mimic animal morphologies and controls. Their body is constrained to produce only legged robots. The control is performed by a Central Pattern Generators network. The specific design of the robots constrains the search space and produce more easily realistic robots, especially robots using legged locomotion.