Adaptive Control of Lower-Limb Exoskeletons for Walking Assistance Based on Inter-Joint Coordination

Abstract

Unilateral motor impairment can disrupt the coordination between the joints, impeding the patient’s normal gait. To assist such patients to walk normally and naturally, an adaptive control algorithm based on inter-joint coordination was proposed in this work for lower-limb exoskeletons. The control strategy can generate the reference trajectory of the affected leg in real time based on a motion coordination model between the joints, and adopt an adaptive controller with virtual windows to track the reference trajectory. Long Short-Term Memory (LSTM) network was also adopted to establish the coordination model between the joints of both lower limbs, which was optimized by preprocessing angle information and adding gait phase information. In the adaptive controller, the virtual windows were symmetrically distributed around the reference trajectory, and its width was adjusted according to the gait phase of the auxiliary leg. In addition, the impedance parameters of the controller were updated online to match the motion capacity of the affected leg based on the spatiotemporal symmetry factors between the bilateral gaits. The LSTM coordination model demonstrated good accuracy and generality in the gait database of seven individuals, with an average root mean square error of 3.5\(^\circ\) and 4.1\(^\circ\) for the hip and knee joint angle estimation, respectively. To further evaluate the control algorithm, four healthy subjects walked wearing the exoskeleton while additional weights were added around the ankle joint to simulate an asymmetric gait. From the experimental results, it was shown that the algorithm improved the gait symmetry of the subjects to a normal level while exhibiting great adaptability to different subjects.