A self-tuning multi-phase CPG enabling the snake robot to adapt to environments

Abstract

Making biomimetic robots move like natural animals is an interesting problem, because this topic involves not only the low level algorithm that controls the movement of robots' bodies and limbs but also the high level control strategy that deals with different kinds of situations. Based on a certain biological assumption, a self-tuning multi-phase CPG for snake robots is proposed. This method imitates the control strategy of natural snake's movement in different environments, which enables the snake robot to move more quickly and naturally. Through kinematic and dynamic analysis of snake robots, optimal control parameters are chosen for the decision strategy. Due to the intrinsic property of the multi-phase CPG, this model can change the movement patterns and control parameters autonomously according to external information. As a result, such neural control provides a powerful but simple way to self-tune adaptable behaviors in snake robots.