In this work, we are presenting a machine learning multi-model algorithm to design a policy based learning from human demonstration where the policy input represents a state - tuple of predicted generated forces measured from human muscles using sEMG electrodes and human arm endpoint trajectory - while the policy output represents encoded trajectories of position and associated force by extracting patterns from the aforementioned demonstrations (policy input).
Али В. (науч. рук. Колюбин С.А.) EMG-BASED TRAJECTORY PREDICTION FOR SKILL TRANSFER IN ROBOTIC MANIPULATION // Сборник тезисов докладов конгресса молодых ученых. Электронное издание. – СПб: Университет ИТМО, [2022]. URL: https://kmu.itmo.ru/digests/article/8519