Triple-step Nonlinear Controller with MLFNN for a Lower Limb Rehabilitation Robot

Abstract

Passive training is the most fundamental functionality of a lower limb rehabilitation robot (LLRR), and high position tracking accuracy can ensure it is completed satisfactorily. In this paper, a triple-step nonlinear controller with a multi-layer feed-forward neural network (MLFNN) is proposed to improve the tracking accuracy of a LLRR. The triple-step nonlinear controller as the basic controller can guarantee the LLRR follow gait trajectory, and the MLFNN is designed based on a specific objective function to compensate the disturbances and system uncertainties. Experiments are carried out on the LLRR, and the results show that the proposed controller can obtain higher tracking accuracy than the triple-step controller without MLFNN.