Reinforcement Learning for Biped Robot

Animal rhythmic movements such as locomotion are considered to be controlled by neural circuits called central pattern generators (CPGs), which generate oscillatory signals. Motivated by such a biological mechanisms, rhythmic movements controlled by CPG has been studied. As an autonomous learning framework for the CPG controller, we propose an reinforcement learning method , which is called the CPG-actor-critic method. We apply this method to the reinforcement learning for the biped robot. The computer simulation shows that our method is able to train the CPG such that the biped robot walks stably. We also examine the characteristic of this CPG controller.