A Modified Dynamic Movement Primitive Algorithm for Adaptive Gait Control of a Lower Limb Exoskeleton

Abstract

A major challenge in the lower limb exoskeleton for walking assistance is the adaptive gait control. In this article, a modified dynamic movement primitive (DMP) (MDMP) control is proposed to achieve gait adjustment with different assistance levels. This is achieved by inclusion of interaction forces in the formulation of DMP, which enables learning from physical human–robot interaction. A threshold force is introduced accounting for different levels of walking assistance from the exoskeleton. The MDMP is, thus, capable of generating adjustable gait and reshaping trajectories with data from the interaction force sensors. The experiments on five subjects show that the average differences between the human body and the exoskeleton are 4.13° and 1.92° on the hip and knee, respectively, with average interaction forces of 42.54 N and 26.36 N exerted on the subjects' thigh and shank. The results demonstrated that the MDMP method can effectively provide adjustable gait for walking assistance.