Trajectory tracking using online learning LQR with adaptive learning control of a leg-exoskeleton for disorder gait rehabilitation

Precise trajectory tracking of gait pattern under varied load condition is necessary for rehabilitation using leg exoskeleton. In this paper, online iterative learning linear quadratic regulator (OILLQR) with adaptive iterative learning control is proposed to control trajectory tracking of a leg exoskeleton for rehabilitation developed at AIT. The algorithm determines an optimized weight matrix of the conventional LQR controller by using online learning method for every point of the trajectory and also the controller gain of each point. Iterative learning control gain is then optimized according to the optimized gain found from OILLQR algorithm which results in good tracking performance. The proposed control algorithm can significantly shorten the learning time of the system compared to the conventional PID and LQR controller with iterative learning control. Simulation and experimental results confirm short learning time of at least 2 times and good tracking performance of at least 40% reduction in tracking error for the proposed algorithm. An experiment on test subject with the system confirms good tracking performance which is suitable for gait disorder patient.