Each reach requires a fixed amount of total torque, split between the exoskeleton and the patient's own muscles. In fixed-assistance mode the robot always contributes the same slider-set torque, so the patient only ever has to supply the leftover — that leftover never grows, no matter how strong the patient becomes, because the controller never asks for more.
robot = fixedLevel (constant)
patient = 1 − fixedLevel (flat, never challenged)
In assist-as-needed mode the controller measures the patient's current strength every trial and fills only the gap:
robot(t) = max(0, 1 − strength(t))
patient(t) = min(1, strength(t))
Δstrength ∝ effort(t) · (1 − strength(t)) per trial
Because effort itself drives the strength gain, AAN produces a self-reinforcing climb: patient effort rises while robot assistance falls, the two lines crossing once the patient supplies half the work. Fixed assistance keeps both lines flat, so little real neuromuscular re-learning happens even though the movement still "succeeds" every time.
- Fixed torque level — the constant assistance a fixed-mode robot applies regardless of patient performance.
- Session speed — how fast simulated training trials play out.
- Recovery progress score — total strength gained (session end − session start) in each mode, once you've run both.