DEVELOPING A NOVEL ROBOT-AIDED TELEREHABILITATION SYSTEM FOCUSING ON HUMAN UPPER LIMB

Abstract

Stroke remains a leading cause of long-term disability worldwide, placing an unsustainable burden on healthcare systems due to the requirement for high-intensity, one-on-one manual therapy. While robot-assisted rehabilitation offers a solution to deliver precise, repetitive training, its widespread adoption is currently constrained by interface complexity, the necessity for co-located clinical expertise, and the lack of interoperability between disparate robotic systems. This dissertation presents the design, implementation, and validation of a novel Cyber-Physical System (CPS) framework for immersive, robot-agnostic telerehabilitation. The research is grounded in the development of the Smart Robotic Exoskeleton (SREx), a 7-Degree-of-Freedom (DoF) upper-limb device governed by a real-time EtherCAT admittance controller. To enable safe remote operation over the public internet, a "Tri-Node" Teleoperation Topology was architected using Azure Web PubSub. This architecture implements a "Local Authority" safety philosophy, where deterministic hardware constraints (velocity limits, collision prediction) reside at the edge, rendering the patient safe from network latency or jitter. To bridge the spatial and cognitive gaps inherent in telemedicine, the framework introduces two critical interfaces. First, a Mixed Reality (MR) based Digital Twin interface allows remote therapists to manipulate a "Ghost" robot to define Passive Range of Motion (PROM) limits and design functional Active Assist tasks in a shared 3D environment. Second, an Intelligent User Interface (IUI) integrates a Large Language Model (LLM) within a Neuro-Symbolic architecture. This system translates natural-language spoken commands into robot control commands, utilizing a physics-based "Pre-Flight Probe" to ensure that AI-generated trajectories are kinematically safe before execution. Experimental validation with healthy participants demonstrated the system’s ability to maintain stable control loops (≈ 10 Hz) with minimal jitter over commercial broadband.