Development of an End-Effector-Based System for Arm Support and Swing During Rehabilitation of Walking.

The theory of interlimb neural coupling suggests that arm swing should be integrated into gait training to promote neural control of walking. We previously developed a mobile gait trainer which followed the user during free overground walking. However, it did not provide assistance in the synchronised arm and leg movements. This work aimed to develop a robotic arm module for the mobile gait trainer which offers stiff vertical arm support for balance, provides adaptable arm movement guidance during assisted walking, and allows free arm swing as experienced in daily walking. Based on an analysis of the wrist movement from normal gait, an end-effector-based system was developed, which used a screw drive to produce the vertical wrist movement and a timing belt drive with a force sensor to control the anterior-posterior wrist movement. Position controllers were employed to generate arm swing in passive training. Admittance algorithms were developed for the timing belt drive to generate adaptable horizontal movement. Free movement was produced by simplifying the admittance algorithms to include only a damping term. Preliminary tests showed that the position control algorithms produced arm swing in passive training with an accuracy of $\mathbf{1. 7 8}$ mm vertically and $\mathbf{3. 1 7 ~ m m}$ horizontally. The admittance controller produced adaptable anterior-posterior movement, where the active force modified the range of swing amplitude. The simplified admittance controller enabled the user to move freely with negligible interference from the robotic arm module. The system managed to offer rigid support for balance, to provide assisted arm swing for gait training with adaptable anterior-posterior movement, and to allow free arm swing. Integration of the arm movement with the mobile gait trainer and mechanism for leg movement will be investigated in future work.