This paper introduces a progressive assist-as-needed (pAAN) controller into our custom-made lower limb exoskeleton system. This control strategy can enhance the active participation of subjects. The controller can dynamically estimate a subject's input (voluntary joint torque) based on electromyography (EMG) without calibrations. The EMG-torque relationship learning is unsupervised. The zero-error estimation of the subject's input is guaranteed by a progressive learning strategy. The adaptive controller adjusts the control inputs of motors to complete predefined trajectories. Online torque estimation and adaptive motion control are both realised in the pAAN controller. Additionally, some practical problems of EMG application, caused by time-varying property of EMG signals and electrode displacement, would be avoided. From the simulation and experimental studies, our pAAN controller can predict the subject's input well, and the exoskeleton helps subjects move precisely. Active participation of subjects is achieved during training.
Gui, K., Tan, U-X., Liu, H., & Zhang, D. (2020). Electromyography-Driven Progressive Assist-as-Needed Control for Lower Limb Exoskeleton. IEEE Transactions on Medical Robotics and Bionics, 2(1), 50-58. https://doi.org/10.1109/TMRB.2020.2970222