Myoelectric Control of a Robotic Exoskeleton for Rehabilitation

Abstract

A primary challenge in the design of human-robot interfaces for rehabilitation after neurological injury, such as stroke or spinal cord injury, is the detection of user intent, needed to maximize the efficacy of the therapy. Common approaches to rehabilitation robot interfaces, including the current implementation of the MAHI Exo-II upper extremity therapeutic exoskeleton at Rice University, rely on impedance control schemes. Another approach, surface electromyography (sEMG), is gaining attention. This interface is appealing as the recorded signal is related to the individual's desired torque about the joint the muscle actuates. In this thesis, an sEMG interface and associated control schemes are proposed and investigated for the MAHI Exo-II. A known drawback of sEMG interfaces are lengthy subject- and session-dependent calibration procedures to develop muscle-force mappings. In this thesis, a relaxed calibration procedure and various control schemes are proposed to enable practical integration into therapy protocols. Agonist-antagonist muscle groups were related following normalization based on sub-maximal isometric contraction in the exoskeleton. Pilot experiments were conducted on healthy subjects to assess the usability of the exoskeleton in the proposed control modes of operation given simple sEMG interface. The results of these experiments support the implementation of the proposed sEMG interface. Future experiments will focus on validation in impaired populations.