Development of Soft Gripper Pneumatic Control System Based on Deep Reinforcement Learning

Access full-text files



Journal Title

Journal ISSN

Volume Title


University of Texas at Austin


As interest in soft grippers soared, many studies have been performed to control the soft gripper. For the soft gripper control, a soft gripper model is required first. Usually, the soft gripper modeling has been done through finite element analysis, which takes lots of time and is effective only in limited situations. Therefore, research on deep learning-based modeling with a small amount of FEM results has been extensively conducted, and some satisfactory results have been reported. However, since the model is expressed in the form of a neural network, it is difficult to utilize general control methods, so research on optimal control or deep reinforcement learning is being attempted. In this study, we propose a pneumatic control system for the soft gripper control based on the DRL. To this end, the soft gripper and DRL-based controller are directly developed, and experiments are performed and the results are analyzed.


LCSH Subject Headings