Model identification and adaptive control of lower limb exoskeleton based on neighborhood field optimization

Abstract

In lower limb exoskeletons, control performance and system stability of human–robot coordinated movement are often hampered by some model parametric uncertainties. To address this problem, Neighborhood Field Optimization (NFO) is proposed to identify the unknown model parameters of an exoskeleton for the model-based controller design. The excitation trajectory is designed by the NFO algorithm with motion constraints to improve the model identification accuracy. Meanwhile, the Huber fitness function is adopted to suppress the influence of the disturbance points in sampled dataset. Then an adaptive backstepping control scheme is constructed to improve the dynamic tracking performance of human–robot training mode in the presence of the identification error. Via Lyapunov technique and backstepping iteration, all the system state errors of the exoskeleton are bound and converge to zero neighborhood based on the assumption of bounded identified parameter error. Finally, the model identification results and comparative tracking performance of the proposed scheme are verified by an experimental platform of Two-degrees of freedom (DOF) lower limb exoskeleton with human–robot cooperative motion.