Gait-phase-adaptive control of lower-limb exoskeletons for motion assistance

Abstract

This study aims to develop a phased control strategy for a hydraulic lower-limb exoskeleton that incorporates motion intention prediction to provide effective and stable walking assistance for users. An inertial measurement unit (IMU)-based long short-term memory (LSTM) network was developed to predict lower-limb motion intention, which was then used as the motion input to the phased control strategy. Impedance control was applied during the support phase, whereas force closed-loop control with position compensation was employed during the swing phase. Comparative walking tests were conducted using joint angles measured by the IMU and motion intention predicted by the LSTM network as the respective inputs. Compared with IMU-measured joint angles, the use of LSTM-predicted motion intention reduced human–robot interaction forces and joint angle tracking errors. Muscle activity analysis of the rectus femoris, biceps femoris and gastrocnemius further showed that the exoskeleton reduced human muscle activation by up to 37.5% during walking. The proposed framework integrates motion intention prediction into a phased control strategy for a hydraulic lower-limb exoskeleton, while individualized impedance parameter tuning is introduced to accommodate differences among subjects. The stability of the control switching process is further theoretically analyzed using a multiple lyapunov functions (MLF)-based approach, strengthening the theoretical foundation of the proposed control framework.