Open-Loop Motion Control of a Hydraulic Soft Robotic Arm Using Deep Reinforcement Learning

Soft robotic arms are of great interests in recent years, but it is challenging to perform effective control due to their strongly nonlinear characteristics. This work develops a model-free open-loop control method for a hydraulic soft robotic arm in spatial motion. A control policy based on reinforcement learning technique is proposed by using Deep Deterministic Policy Gradient. The kinematic model of the soft robotic arm is employed instead of physical prototype to train the control policy. A complete training framework is established through the Reinforcement Learning Toolbox and Deep Learning Toolbox in Matlab software. To make the control policy fast converge and avoid falling into local optimum, the reward is shaped by combining the position error and the action together. A series of simulations are implemented and the results verify the effectiveness of the control policy. It is also shown that the proposed control policy can achieve both of good stability and tracking performance simultaneously.