Learning to Suppress Tremors: A Deep Reinforcement Learning-Enabled Soft Exoskeleton for Parkinson's Patients

Abstract Neurological tremors, prevalent among a large population, are one of the most rampant movement disorders. While biomechanical loading and exoskeletons show promise in enhancing patient well-being, traditional control algorithms limit their efficacy in dynamic movements and personalized interventions. Furthermore, there exists a pressing need for more comprehensive and robust validation methods to ensure the effectiveness and generalizability of proposed solutions. This paper proposes a physical simulation approach modelling multiple arm joints and tremor propagation. This study also introduces a novel adaptable reinforcement learning environment tailored for disorders with tremors. Furthermore we present a deep reinforcement learning based encoder-actor controller for Parkinson's tremor present in various shoulder and elbow joint axes, displayed in dynamic movements. We found that the controller is capable of suppressing tremors in all recorded dynamic movements. The control achieves an average tremor of over 57.43% for three out of the four distinct movement trajectories. Notably, the controller attains a maximum accuracy of 99.49%. Our findings suggest that a deep reinforcement learning based control strategy offers a viable solution for tremor suppression in real-world scenarios. By overcoming the limitations of traditional control algorithms, this work takes a new step in the adaptation of biomechanical loading into the everyday life of patients. This work also opens avenues for more adaptive and personalized interventions in managing movement disorders.