Model Free DEAP Controller Learned by Reinforcement Learning DDPG Algorithm

In this work the Reinforcement Learning method is applied to construct a controller for Dielectric Electroactive Polymer actuator. The learning of control policy is model free without prior knowledge about the exact model of device dynamics. Due to the nonlinearities, multiple time constants and output feedback the control design is challenging task. To solve the problem, the recent Deep Deterministic Policy Gradient is applied. Further, the controller is designed with and without model reference.