Design, Development and Validation of New Control Strategies for Robot-Assisted Hand Rehabilitation

Wearable robotics has proven to be very effective in rehabilitation scenarios. This thesis presents the design and evaluation of a series of control strategies and algorithm for hand exoskeletons developed to use in rehabilitation scenarios. Three devices were used in this work, HandeXos-beta, HandeXos-gamma, and HABILIS. Each of these features a SEA architecture. A thorough low-level control characterization was performed on all aforementioned devices. High-level control algorithms were developed: a user-intention detection routine was developed and implemented on HX-gamma, based on data extracted from the kinematic variables recorded by the exoskeleton and an additional sEMG sensory system. The algorithm features a machine learning component and was implemented in a simulation to assess its performance in detecting user intention and classifying the intended grasp motion. The HABILIS device was fitted with a high-level exercise library with a finite-state machine algorithm. Moreover, as the device features a series of linear springs to enact extention torques,a compensation algorithm to nullify the effect of these return springs was implemented. Finally, the control algorthm developed on these devices were adapted and assessed on different robotic exoskeletons to gauge their effectiveness in different scenarios. In particular, the control strategies presented were assessed on a shoulder-elbow upper limb exoskeleton (NESM) and on a lower-limb exoskeleton for gait assistance (APO).