Design and State Machine Control of an Exoskeleton Frame Integrated to Telerehabilitation Network for Duchenne...

Exoskeletons are valuable tools for the physical rehabilitation of individuals with paralysis or weakness. Exoskeleton designs differentiate to meet the specific needs of medical conditions. However, there is limited research on wheelchair-mounted exoskeletons for patients who are unable to move overall but have greater upper extremity needs. This study presents the electromechanical design of a wheelchair-mounted exoskeleton for Duchenne Muscular Dystrophy (DMD) patients and the use of finite state machine (FSM) state classification for its control. The exoskeleton features four degrees of freedom and incorporates Inertial Measurement Units (IMU), Electromyography (EMG), and Force Sensitive Resistor (FSR) sensors. The FSM states and transitions are defined for controlling the exoskeleton, but instead of relying on traditional rule-based methods, 1D-CNN models are used to predict the current state and trigger transitions when state changes occur. Furthermore, to ensure the exoskeleton only provides assistance when needed, it activates based on FSR data indicating rest while the system predicts motion. The shallow models used in the experiments achieved 84% accuracy in classifying FSM states, demonstrating that machine learning models can be effectively applied for FSM control. By replacing the decision-making component with a common machine learning model, the development of such systems becomes significantly easier and more adaptable to the specific needs of patients.