FIGURE 12.

Case3: Performance of the RL based controller under human-exoskeleton interactions during more squatting cycles. (A) Human-exoskeleton interaction (strap) forces during multiple squatting cycles. hip$–$force, femur$–$force and tibia$–$force in the figure represent the interaction forces between the human and the strap on the exoskeleton, respectively. (B) Performance of the RL based controller. The first figure demonstrates the foot CoP reward (green and red lines), end-effector reward (magenta line) and joint position tracking reward (black line) calculated according to Eqs 10–15. The bottom figure depicts the actions for the hip, knee ankle joint predicted from the neural network.