A hybrid GP/GA approach for co-evolving controllers and robotbodies to achieve fitness-specified tasks

摘要

Evolutionary approaches have been advocated to automate robotdesign. Some research work has shown the success of evolving controllersfor the robots by genetic approaches. As we can observe, however, notonly the controller but also the robot body itself can affect thebehavior of the robot in a robot system. We develop a hybrid GP/GAapproach to evolve both controllers and robot bodies to achievebehavior-specified tasks. In order to assess the performance of thedeveloped approach, it is used to evolve a simulated agent, with its owncontroller and body, to do obstacle avoidance in the simulatedenvironment. Experimental results show the promise of this work. Inaddition, the importance of co-evolving controllers and robot bodies isanalyzed and discussed