Abstract:
Neurological injuries account for one of the leading causes of physical disability. Physiotherapy has been shown to be an effective means of restoring function to survivors of such injuries. However, conventional therapy can be tedious and subjective. Over the last few decades, considerable research has been conducted on the use of robotics to supplement therapy. While it has shown promise, uptake remains low, with current devices tending to be too expensive and complex for practical use in clinical settings. This thesis presents a new robotic component which is a type of elastomeric series elastic actuator (eSEA). The aim is to improve rehabilitation robotics though simpler design, modularity and better physical human-robot interaction (pHRI). SEAs, which place a spring between an actuator and its load, are an effective means of achieving safe and comfortable pHRI. eSEAs build on this concept by using materials with inherent damping, which can further improve pHRI, increase force-volume ratios and improve safety. The first contribution of the thesis is the derivation of a model that allows the eSEA to measure pHRI torque. It is validated over the course of 8-9 h of elastomer compressions, where it is shown to reliably measure torque with high accuracy. It is then inverted in preparation for model-based torque control with the eSEA. Sliding mode control (SMC) and model predictive control (MPC) are compared for pHRI torque control at the elbow with an exoskeleton utilizing the eSEA. MPC provided more accurate torque control for well-known dynamics, however, the robustness of SMC better compensated for uncertain human dynamics. The next chapter of the thesis further explores topics of pHRI by quantitatively and qualitatively comparing motion smoothness and human perception of interaction with an elastomer to that with a spring. More research is needed to obtain conclusive results, however, the evidence suggests that the elastomer provided more natural and smooth pHRI, in addition to being perceived more favorably. The final chapters of the thesis discuss applications of the eSEA in applications related to human-robot interaction and therapy. The eSEA is first used in a study that virtually normalizes physical impairment, to create a tool to study motor learning. The second application details a framework that combines clinician expertise with robotic measurements. Each application deploys the eSEA at a different biological joint, displaying its modularity and versatility. These applications also illustrate how the eSEA can be used to enhance the capabilities of clinicians, ultimately leading to better outcomes for survivors of neurological injuries.