Adaptive Admittance Control Strategy for a Robotic Knee Exoskeleton With a Nonlinear Variable Stiffness Actuator

Abstract

This article presents the design and control of a robotic knee exoskeleton for gait rehabilitation of patients with knee joint impairments. First, the hardware design of the exoskeleton is presented, including the mechanical structure, actuator design and configuration, and electronic system. Based on the nonlinear characteristics of human muscles, a nonlinear variable stiffness actuator (NLVSA) is designed for the actuation system of the exoskeleton. Next, the modeling of the NLVSA is described. In addition, an adaptive admittance control strategy comprising a sparrow search optimization algorithm-based long short-term memory neural network model and an adaptive admittance control algorithm based on the radial basis function neural network (RBFAAC) is proposed for the exoskeleton. Finally, a pilot study is conducted to demonstrate the effectiveness of the robotic knee exoskeleton. The experimental results validate the effectiveness of the designed NLVSA, and the exoskeleton has the potential for human knee rehabilitation by providing effective assistance with the proposed control strategy. With the proposed RBFAAC algorithm, the average root mean square error between the reference and actual knee joint angles is 1.24° at different walking speeds.