Shape Memory Alloy Driven Soft Robot Design and Position Control Using Continuous Reinforcement Learning

Abstract

In this paper, we introduce a new Shape Memory Alloy (SMA) driven biomimetic soft robot and developed its mathematical model of a spherical coordinate system. Aiming at the problem of the control of the soft robot, a new reinforcement learning (RL) based position control algorithm, Soft robot Position Control Deep Deterministic Policy Gradient (SPCDDPG), is proposed. Our approach enables the soft robot to learn the target reaching task by performing continuous actions without any prior knowledge of the configuration. Moreover, we developed a simulation system for learning the target reaching task and designed a reward function for the system. Finally, the experimental results demonstrate the effectiveness of the proposed method which is tested and validated through extensive numerical simulations under different conditions.