A Versatile Upper-Limb Exoskeleton for Neurorehabilitation

Lesions of the central nervous system are one of the major causes of disability in the world. Many patients suffering from an impairment of neuromotor functionality have the potential to recover at least some of their movement ability by harnessing the principle of brain plasticity. Intensive and individualized neurotherapy can substantially induce plasticity leading to a more complete recovery compared to relying only on spontaneous neural processes. Providing a sufficient amount of intensive and individualized therapy to patients is an ever-growing challenge. Financial restrictions of the healthcare system and the availability of skilled and experienced therapists put bounds on the time therapists can spend on treating a patient. The steady growth of demographic aging, combined with the high incidence of stroke in the elderly population, will accelerate the urgency of this challenge in the near future. Thus, finding solutions to provide therapy with better efficacy of the therapist's work time has a substantial impact on recovery from disability. Robot-assisted therapy offers standardized and data-based intervention while allowing therapists to use their work time more effectively by automating some aspects of therapy. Particularly, robots that can guide, assist, and resist movements of the full limb have the potential to substitute a substantial part of the therapist's interactions for any state of the patient's impairment severity (e.g., severely to mildly affected). Thus, a broad adoption of therapy robots in clinics would promise the biggest impact of these benefits. Fully assistive state-of-the-art therapy robots (i.e., exoskeletons like ArmeoPower, Hocoma AG, Switzerland) were so far mostly restricted to application with severely affected patients. Their restricted range of motion, speed, and haptic transparency limited the applicability, particularly for moderately to mildly affected patients. Partially assistive robots like Diego (Tyromotion, Austria) could not correct or assist posture and were restricted in rendering exercises with selective resistance or constraints. Thus, the overall goal of this thesis was to develop a fully assistive robotic system (i.e., exoskeleton) that offers the technical capability to provide meaningful therapy for the full ability range of patients, i.e., severely to mildly affected. The main contributions of this thesis were: Kinematic Design: We developed a kinematic structure for an exoskeleton involving all relevant degrees of freedom of the human arm. Particularly, the kinematic design of the spherical robotic glenohumeral and wrist joint combined with controlled shoulder coupling maximized the manipulability while providing almost the full active range of motion of humans. Thereby, we could enable unprecedented freedom of movement while being assisted by the robot, which is essential particularly for moderately to mildly affected patients. Hardware Design: We proposed a linkage and human-robot attachment design together with a selection of actuators and sensors, equipping the robot with sufficient strength, speed, and haptic rendering capability while preventing collisions with the human while traveling through the range of motion. Thereby, the robot system can provide a versatile range of interactions with the user, \eg haptic transparency, fully guided and assisted movements, heavy resistance for maximum strength training, and relatively stiff haptic walls. Arm and Torso Attachment: We introduced a new concept for the human-robot attachment of the arm and a simple but effective haptic reference for the upper torso. In a case study of four participants, we could demonstrate the improved resilience of joint alignment under load and improved attachment repeatability of this proposed concept compared to conventional solutions. Scalupa Support: We developed a body-powered orthosis that could stabilize and mobilize the scapula on the physiologic scapulohumeral rhythm. In a case study of four participants, the slightly modified orthosis was combined with our new exoskeleton, demonstrating effective support of the scapula during the use of the robot. Thus, we were the first to find a solution on how to support the scapula without therapist intervention during robot-assisted therapy. Thereby, we could increase the potential autonomy of rehabilitation exoskeletons in treating patients with distorted scapular stability and mobility. Control Modularity: We adopted control algorithms for modular task organization and optimization from the field of autonomous robots to the rehabilitation device. We could successfully demonstrate merging tasks defined in isolated software modules to a desired overall system behavior. Thus, we proposed and validated a method facilitating the implementation of the control policies for this complex robot in modular junks defined in the most convenient and intuitive representation space. High-Fidelity Haptic Rendering: We proposed an easy-to-tune control policy for robust interaction force control reducing the tracking error substantially in the presence of nimble movements dictated by the user. Furthermore, we demonstrated the combination of this method with kinematic constraints and the controlled shoulder coupling using the above-mentioned optimization method. Thus, we introduced a control system for wide impedance range haptic rendering. The ANYexo 2.0, the first device to unite these capabilities, could be the vanguard for highly versatile robotic therapy assistants applicable to a broad target group and an extensive range of exercises. Thus, this work is a step in advancing the practical adoption of these devices in clinics