Nonlinear Learning Control of Ionic Polymer Metal Composites

Abstract Ionic polymer-metal composites (IPMCs) are promising smart materials for various applications including micro aero vehicles (MAVs), underwater vehicles, robot engineering, and medical devices. However, characteristics of IPMCs vary with various factors including temperature, hydration, surface treatments and inter-layer conditions. Moreover, they show nonlinear behaviors such as hysteresis and back relaxation. In order to overcome these challenges for precise control of IPMCs, we employ a probabilistic learning method. Transient time response of the IPMC tip deflection is modeled with probabilistic parameters, and Bouc-Wen model is added to represent hysteresis with respect to actuation voltage. A controller is designed based on a reinforcement learning algorithm that can address probabilistic uncertainties and nonlinearity while keeping the simplicity of PID control that can be easily implemented and has advantages from a practical viewpoint. Simulations and experiments show the reliable performance of the proposed scheme against the uncertainty and variations of the parameters.