Towards Efficient Personalization of Computerized Lower Limb Prostheses Via Reinforcement Learning in a Clinical Setup - Group 1

Résumé succinct

The goal of this clinical trial is to understand the feasibility and effectiveness of using reinforcement learning to personalize robotic prosthetic legs (an experimental prototype) for unilateral transfemoral amputees. The main questions it aims to answer are: * With the developed RL-based Recommendation Interfacing System (RISE), clinicians are able to personalize prosthetic legs faster compared with existing manual personalization procedures. * With the developed RL-based Recommendation Interfacing System (RISE), clinicians are able to personalize prosthetic legs without detailed knowledge about how the prosthetic legs are controlled. * Patients perform better when the prosthetic legs are personalized with RISE system compared with the ones personalized manually Researchers will compare two arms (RISE guided personalization and manual personalization) to see if the tuning speed will increase and if patients can perform better. Participants will go through the standard prosthetic fitting procedures, such as alignment adjustment, then they will experience repeated prosthesis personalization procedures conducted by tuning specialists without RISE, tuning specialists with RISE, and prosthetists (without tuning expertise) with RISE on different types of terrains. In the end, the participants will go through a testing trial, in which they will experience the prototype personalized through the three different approaches without knowing how the control parameters are decided. Their walking performance will be recorded. It is expected that the participants will visit the testing site 8 times, which including alignment (1 visit), three personalization procedures (twice for each), and one testing trial (1-2 visits).