Closed-Loop Control of Large Assistive Torques to Unknown External Loads via a Back Support Exoskeleton

Most control strategies for active back-support exoskeletons (BSEs) do not target users' internal forces while providing assistance, limiting synergy with the user. In our previous study, we proposed a neuromechanical model-based controller (NMBC) that directly targeted lumbosacral (L5/S1) active moments. Despite promising results, the use of a soft exosuit limited the maximum assistance possible, leaving unclear whether increased assistance would have more benefits for the user. Therefore, in this study, we adapted the NMBC to provide assistance via an active rigid BSE and were able to provide higher assistive torques, up to 114.5 Nm. We tested three support levels ($20 \%, 50 \%, 80 \%$ of estimated active L5 / S1 moments) provided to three participants lifting two loads (5 kg, 15 kg) and two lifting techniques (stoop, squat) and compared it with transparent mode (MINIMP) and without exoskeleton (NOEXO). All assisted conditions significantly reduced L5/S1 moments compared to NOEXO, despite the BSE being heavy. A general trend towards reduced effort was observed with higher gains, although, not statistically significant. However, large assistive torques also caused deteriorated torque tracking and instabilities. The system adapted to unknown loads for both techniques. This study demonstrated for the first time the feasibility of using NMBC with larger support gains, highlighting both its potential benefits and challenges.