2021/03/04 by Nat Wannawas, Wannawas, Nat, Mahendran Subramanian +3
Engineering · Neuroscience · #EEG and Brain-Computer Interfaces #FOS: Computer and information sciences #Machine Learning (cs.LG) #Muscle activation and electromyography studies #Neuroscience and Neural Engineering
paper · pdf · doi:10.48550/arxiv.2103.03057
openalex publication_date 2021/03/04 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Functional Electrical Stimulation (FES) can restore motion to a paralysed\nperson's muscles. Yet, control stimulating many muscles to restore the\npractical function of entire limbs is an unsolved problem. Current\nneurostimulation engineering still relies on 20th Century control approaches\nand correspondingly shows only modest results that require daily tinkering to\noperate at all. Here, we present our state of the art Deep Reinforcement\nLearning (RL) developed for real time adaptive neurostimulation of paralysed\nlegs for FES cycling. Core to our approach is the integration of a personalised\nneuromechanical component into our reinforcement learning framework that allows\nus to train the model efficiently without demanding extended training sessions\nwith the patient and working out of the box. Our neuromechanical component\nincludes merges musculoskeletal models of muscle and or tendon function and a\nmultistate model of muscle fatigue, to render the neurostimulation responsive\nto a paraplegic's cyclist instantaneous muscle capacity. Our RL approach\noutperforms PID and Fuzzy Logic controllers in accuracy and performance.\nCrucially, our system learned to stimulate a cyclist's legs from ramping up\nspeed at the start to maintaining a high cadence in steady state racing as the\nmuscles fatigue. A part of our RL neurostimulation system has been successfully\ndeployed at the Cybathlon 2020 bionic Olympics in the FES discipline with our\nparaplegic cyclist winning the Silver medal among 9 competing teams.\n