Development and experimental testing of a control strategy for adaptive assistance delivered by a semi-passive upper-limb exoskeleton.

Work-related musculoskeletal disorders (WMSDs) are one of the most common causes of occupational disease in Europe. They typically cause muscle fatigue and injuries, leading to a relatively high number of lost workdays with a negative impact on enterprises’ productivity and the increased burden on national healthcare systems, as reported by the EU-OSHA in 2007. In recent years, in order to reduce workers’ fatigue and the insurgence of WMSDs, a growing interest in wearable robotics (WR) technologies has led to the application of upper-limb exoskeletons for industrial applications. The state of the art provides many examples of passive devices where the assistive action (e.g. assistive torque) is generally achieved by means of passive mechanisms, like springs and dampers. Despite their light weight and potentialities in reducing workers’ muscular strain, they still have the intrinsic lack of adaptivity to different tasks (e.g. static or dynamic) or different physiological state of their users (e.g. increased fatigue). Indeed, once the support level is selected at the beginning of the work it remains fixed at a certain level for the whole duration of the task, unless the worker manually adjusts it. At the leading edge of this research field, the WR Laboratory of Scuola Superiore Sant’Anna of Pisa is currently investigating innovative solutions. This thesis, developed within the WR Laboratory, aims at designing, developing, and eventually implementing an adaptive algorithm for the automatic selection of the support level for a semi-passive upper-limb exoskeleton (H-PULSE). The exoskeleton was developed by IUVO S.r.l., a spin-off company of Scuola Superiore Sant’Anna, within the H2020 HUMAN project (Grant Agreement n. 723737, website: http://humanmanufacturing.eu). The H-PULSE is embedded with sensors, to monitor the shoulder flexion/extension angle, and an active tuning mechanism that allows to automatically change, in real time, the pre-tensioning of the springs responsible for the assistive torque peak. The algorithm developed in this work of thesis exploits shoulder kinematics to compute movement-related features to online estimate the desired level of support. In particular, the movements features are linearly combined in a finite different equation formulation to compute the level of assistance that the exoskeleton needs to provide. The algorithm was first designed and simulated offline in MATLAB R2019b and then translated in LabView 2018 for evaluating its real-time performance. Finally, the algorithm was implemented in the real-time processor of the H-PULSE exoskeleton for experimental testing. The algorithm was evaluated with human-in-the-loop experimental trials, with the main objectives of investigating the effects of the adaptive control strategy against experimental conditions where the exoskeleton was not worn (i.e. without any support) or delivered a fixed and predetermined level of assistance (i.e. supporting about the 50% of the arm gravitational torque).