Analyse des performances en propulsion d’utilisateurs de fauteuil roulant

Abstract

Nearly 200,000 Canadians use a manual wheelchair (MWC), an activity that places high mechanical demands on the upper limbs and is associated with an increased risk of shoulder pain and injuries. Propulsion training therefore represents a major challenge in rehabilitation, yet it remains poorly structured. This thesis aimed to assess the impact of a personalized training program, generated in real time by a reinforcement learning algorithm, on the biomechanics of MWC propulsion. The experiment was conducted on a simulator developed at the Laboratoire d’Innovation Ouverte en Technologies de la Santé (LIO-ÉTS) and involved 20 healthy, inexperienced participants. Six training conditions were tested, combining resistive or assistive–resistive feedback under three objectives: increasing the mechanical effective force (MEF), reducing the physiological cost, or combining both criteria. Results showed that resistive training increased mean MEF by more than 6% across the cohort on both sides, at the expense of a physiological cost (+25% on the left and +17% on the right). Assistive-resistive training, on the other hand, reduced the mean physiological cost by up to -34% compared to the initial condition. These findings partially confirm the effectiveness of the personalized feedback model, capable of adjusting task difficulty to each individual. In addition, a control-inspired identification method enabled the estimation of nine intrinsic and reflexive parameters characterizing upper-limb dynamics. The future integration of these parameters into learning algorithms paves the way for adaptive and safe robotic rehabilitation training.

Date	12 Nov 2025
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Rachid Aissaoui (Supervisor) & Sylvie Nadeau (Co-supervisor)