HUMAN-ROBOT INTERACTION: INVESTIGATING CONTROL PARADIGMS FOR END-EFFECTOR BASED UPPER LIMB REHABILITATION

Abstract

Various robot-mediated approaches for upper-limb therapy with encouraging results have been suggested. Yet, owing to mixed results, it is not clear how the interaction control between an end-effector based robot and a user should be ideally designed. In this thesis, we first show that the motor performance of stroke patients saturates in presence of high assistance. Building on this knowledge, an adaptive controller that does not constrain movement variability and incorporates the challenge point framework was developed. The evaluation with stroke patients showed clear task performance improvements and reductions of impairment. Moreover, the inherent factor of social interaction in traditional therapy was considered by investigating human-human interactions of differently skilled partners. Results suggest that learning with a peer is preferable over the interaction with an expert. The results of this thesis provide a further understanding of how the interaction control should be designed to complement the therapist’s work.