An evolutionary system for automatic robot design

Abstract

Designing a robot is generally considered to be demanding, because the designer not only has to predict the interaction between the robot and its environment, but also has to deal with the ensuing problems. The paper examines the way to use evolutionary techniques to synthesize robots automatically. The explorations range from evolving simple behavior controllers for robots, to complex behavior controllers, and finally to accomplishing a complete robot system including its controller and physical structure. Our evolutionary framework includes two subsystems: the genetic programming part that evolves the controller; and the genetic algorithm part for the robot body. We conducted a series of experiments to evaluate our evolutionary framework and the results show that our system can be used to reduce the load of a robot programmer.