A Predictive Control for Pneumatic Muscle Actuators based Exoskeleton by Using MIMO Echo State Network

In this paper, a tracking control for the Pneumatic Muscle Actuators (PMAs) based rehabilitation exoskeleton by using multiple-input-multiple-output (MIMO) Echo State Network (ESN) is proposed. Due to the intrinsic features of the PMAs based exoskeleton (nonlinearities, unmodelled uncertainties, time-varying parameters, hysteresis, etc.), the system is difficult to be modeled accurately. Hence the MIMO ESN is used to approximate the dynamical model of the PMA-driven exoskeleton with a nonlinear autoregressive exogenous model. A single layer network (SLN) is constructed to solve quadratic programming problem over a finite horizon. Control signals are computed according to the trained SLN. The proposed control strategy turns out to be asymptotically stable when the MIMO ESN is capable of approximating the dynamics of the PMAs-driven exoskeleton. Experiments were conducted, and the corresponding results indicated the performance of the proposed strategy is better than that of the traditional approach in trajectory tracking tasks.