An efficient MPC controller for the rehabilitation of an actuated knee joint orthosis

The present paper deals with the position control law of a knee-orthosis acting at the knee joint level for a passive, resistive, and assistive rehabilitation purpose. The nonlinear dynamic model of this lower limb exoskeleton is linearized and decoupled by feedback linearization technique, and a linear Model Predictive Control (MPC) is designed. Then a comparative study between MPC, Constrained MPC (CMPC), Nonlinear MPC (NMPC), and PID highlights the limits of this latter especially in rejecting the disturbance and also in violating the input constraints when parameters variation of the model is studied. We showed that CMPC and NMPC succeed to conceal this violation. Different simulation scenarios for the knee flexion-extension prove the effectiveness of the model predictive approaches.