Acquisition of Goal-Oriented Behavior for Snake-Like Robot by CPG and Reinforcement Learning

Abstract

In this study, we assembly the snake-like robot using ROBOTIS Bioloid and then we realize that the snake-like robot acquires the goal-oriented behavior by CPG and reinforcement learning. CPG (Central Pattern Generator) is a system generating the periodical motion patterns. We apply reinforcement learning to the snake-like robot in order to optimize the offset values of motors. The framework using CPG and reinforcement learning enables the acquisition of behavior knowledge based on physical characteristic of the assembled snake-like robot. Moreover, we realize the autonomous snake-like robot with Raspberry Pi and Web camera. The autonomous snake-like robot by itself can acquire the goal-oriented behavior.