Reinforcement Learning-Based Minimum Energy Position Control of Dielectric Elastomer Actuators

This article deals with the closed-loop optimal control of mechatronic devices based on dielectric elastomer membranes. The goal is to minimize the input electrical energy required to achieve a given position regulation task. The actuator is modeled based on a free-energy framework, which provides a thermodynamically consistent characterization of the losses that occur during actuation. Due to the strongly nonlinear behavior of both system model and dissipation function, traditional techniques based on the analytical solution of the Hamilton-Jacobi-Bellman (HJB) equation cannot be applied. Therefore, a reinforcement learning-based algorithm is here proposed as a tool to solve, offline, the HJB equation related to the energy minimization problem. After discussing the theory, experimental results are presented to validate the effectiveness of the proposed approach for different positioning tasks.