Iterative learning sliding mode control for output-constrained upper-limb exoskeleton with non-repetitive tasks

Abstract

This paper investigates an iterative learning control scheme for a 5 degrees of freedom upper-limb exoskeleton with time-varying constraints. No priori knowledge of nonlinear dynamic parameters is needed in the modeling process by applying the iterative recursive method, thereby the time-consumption modelling work can be averted. Unlike most iterative learning algorithms that require identical initial condition between different iterations, the estimation errors of learning part in our work can be arbitrary small along the iterations under random initial status. To further improve the safety of human-robot system, a tan-type barrier Lyapunov function is utilized in convergence analysis to guarantee the constraint requirements of tracking errors between human body and robot. Furthermore, in order to handle the nonparametric uncertainties and external disturbances, sliding mode control strategy is constructed where the chattering phenomenon in control torques is suppressed by adaptive method. The stability over the time domain and convergence over the iterative domain are proved by composite energy function. Finally, comparative study and co-simulation results of Matlab/Simulink and ADAMS are presented to verify the advantages and effectiveness of proposed algorithm.