Fig. 4

Snapshots of the walking control of the LLRE. The learned controller trained from the decoupled RL-based control framework enables the LLRE to perform walking assistance under varying human-exoskeleton interactions from human subjects with various neuromuscular disorders. a Autonomous walking control without human involved. b robust walking control with a fully healthy, muscle-actuated human. c robust walking assistance with a human with muscle weakness. d robust walking assistance with a human with left hemiparesis. The color of the muscle indicates its activation, with purple being the highest and blue being the lowest