Fig. 5From: Robust walking control of a lower limb rehabilitation exoskeleton coupled with a musculoskeletal model via deep reinforcement learningStatistical analysis results during 40 walking cycles without human or with human subjects under different neuromuscular conditions (curve: mean; shade: STD). From left to right are exoskeleton joint angles, joint torques, and rewards, including the joint position tracking and end-effector reward in Eq. 10 and the CoP reward in Eq. 12. a without a human; b with a passive human; c with a healthy human; d with a human with muscle weakness; e with a human under the left hemiparesis conditionBack to article page