Design of an adaptive gait trajectory controller based on a hybrid two-legged robot model

Abstract

This paper presents a new algorithm for the automatic adaptation of planar motion for an exoskeleton device. The designed control algorithm aims at adapting the motion of the hemiparetic patient “wearing” the exoskeleton to the desired gait based on the patient's active torque. The patient-exoskeleton coupling is modeled as a stand-alone planar nonlinear hybrid two-legged robot model with point feet. The control algorithm consists of a feedback linearization method extended with an optimal controller that assures the tracking of the reference trajectories and disturbance rejection. This control structure is capsuled by an adaptive iterative learning control, which in every motion cycle adapts the reference joint-angle trajectories for the exoskeleton to follow in the next cycle. Motion captured data from healthy subjects was used as reference input in the closed-loop system, because these trajectories guarantee human-like behavior for the robot model with point feet. Simulation results of the iterative learning controller show promising results with respect to cyclic disturbances, as associated with hemiparetic spasticity.