Development and Usability Evaluation of an Asymmetric Torque-driven Lower-limb Exoskeleton for Hemiparetic Gait Rehabilitation With AI Integration

This study presents the design and control implementation of a lower-limb exoskeleton intended to assist hemiparetic gait by specifically addressing inter-limb asymmetry after stroke. The proposed system combines four active and four passive degrees of freedom, and adopts an asymmetric actuation scheme. System-level functionality and control performance were validated through experimental testing. In addition, an initial usability evaluation was conducted with six post-stroke patients to assess the structural and functional feasibility of the robot in a clinical setting. The results revealded strengths in perceived stability and effectiveness along with needs for improvement in adjustability and perceived weight. Finally, a deep learning-based clustering approach was employed to explore long-term gait recovery trajectories from longitudinal walking data. Rather than serving as a predictive tool, the resulting recovery-pattern clusters are proposed as exploratory reference categories that may inform the future design of rehabilitation control strategies, such as group-informed initialization and progression planning. These results suggest that the proposed system has the potential to support hemiparetic gait rehabilitation by addressing inter-limb asymmetry and enabling personalized assistance strategies.