Fig. 1

A. Experimental protocol: the Enrollment session including initial evaluations and GA without the APO (NoExo) was followed by a Tuning session to select the assistive parameters. Participants then performed a Pre-training Assessment session (PreTA) with two 6mWTs in NoExo and with the APO (Exo) and eight Training sessions in Exo. Last, the Post-training Assessment (PostTA) was identical to the PreTA with a final GA. B. Tuning procedure for two representative participants of the Symmetry (ID2) and Speed (ID4) groups. For the Symmetry group, Step #0 aimed at identifying hip angle abnormalities by comparing hip kinematics with the physiological range (grey area). Then, in Step #1, data from the GA—where the 0% of the gait phase corresponded to heel strike (HS)– were shifted to reset the phase at the hip flexion peak (HFP). Draft assistive torque profiles were designed (dashed lines) to improve temporal symmetry: on the sound side, the flexion peak (black circle) was delivered earlier than the physiological one to promote an earlier flexion and reduce stance time; on the prosthetic side, the extension peak (grey square) was delayed with respect to the physiological pattern to promote a longer stance. During Step #2, each participant walked in TM and AM with the initial torque profiles (colored dashed lines); amplitude was gradually increased based on the participant’s preference and phase and duration were fine-tuned. At the end of the session, the final torque profiles (solid bold colored lines) were locked. For the Speed group, the Tuning procedure started from Step #1 and the draft torque profiles were designed to maximize the net power transferred to the participant: peak torque phases were synchronous to flexion/extension velocity peaks and torque durations corresponded to flexion/extension durations. Step #2 and Step #3 followed the same procedure as for the Symmetry group