Hybrid Filtered Disturbance Observer for Precise Motion Generation of a Powered Exoskeleton

Abstract

Lower-limb exoskeletons are promising applications of robotic rehabilitation for people with motor impairment. As current studies have tailored the design of gait trajectories for the target users, realizing a high-precision motion control is a critical issue for safe and effective assistance. The walking assistance involves unique characteristic phases that embody different physical constraints and requirements for assistance. Conventional methods often utilized gain-switching control for time-varying adaptation. However, despite their intuitiveness as well as simplicity, the control performance was unsatisfying due to unmodeled responses by human behavior and continuous interaction with the external environment. To tackle these challenges, this study proposes a hybrid control method applied to the disturbance observer that can provide robust robotic rehabilitation. The proposed method adaptively identifies the exoskeletal system as a hybrid nominal model and online exchanges model-based tracking controllers parallelly to the gait phase of a user. Furthermore, a unique filter named allowance filter is introduced to compensate for the plant dynamics, preventing instability of the inverted plant and realizing digital implementation. In this article, a practical user with complete paraplegia participated in the experiments for verification of the proposed methods.