Muscle-based skeletal bipedal locomotion using neural evolution

Abstract

The developmental of bipedal walkers has long been a very important problem in the fields of robotics and computer simulation. In this work we propose using evolutionary methods and muscle based control to evolve bipedal walkers. We present an efficient method for evolving walking controller, which are used by a robot to walk large distances. Our approach mimics both the human limb control mechanism and the human learning process to create natural walking patterns and provides valuable insight into the learning process.