Fig. 7.

Examples of therapist inputs, interaction forces, and stride-to-stride adaptation during telerobotic locomotor training. Left Panels: Raw time-series data showing kinematics and forces across a single step for Patients 4 and 6 (P4 and P6). The positions of the manipulandum (dashed black line) and the robotic arm attachment point (solid black line) are shown. The measured interaction force ${\mathit{F}}_{\mathit{R}}$ (black line) reflects the actual force applied to the patient’s leg. The robotic controller tried to make ${\mathit{F}}_{\mathit{R}}$ match the trainer commanded force ${\mathit{F}}_{\mathit{T}}$ (red line), but the aggressiveness of the tracking was modulated according to the gait cycle. During the stance phase, a low-gain was used and was increased six-fold during the swing phase. The blue lines and arrows after heel-strike in the vertical position time-series highlight patient differences in the downward leg velocity at heel-strike; P4 hit the ground harder than P6, creating a sizeable vertical ${\mathit{F}}_{\mathit{R}}$. Note that trainer force transmission was software limited to be in the anterior and vertical directions; the trainer force was zeroed during a time-window starting at heel-strike and ending 200 ms later. Right Panels: Example of stride-to-stride adaptation in Patient 1 (P1). At the start of the trial, P1 walked without assistance (black diamonds), followed by several minutes of walking with assistance from a human trainer using the telerobotics system (red circles; the black trend lines are overlaid for illustrative purposes only)