Learning Ergonomic Control in Human–Robot Symbiotic Walking

This article presents an imitation learning strategy for extracting ergonomically safe control policies in physical human–robot interaction scenarios. The presented approach seeks to proactively reduce the risk of injuries and musculoskeletal disorders by anticipating ergonomic effects a robot's actions on human partner, e.g., how ankle angle prosthesis affects future knee torques user. To this end, we extend ensemble Bayesian primitives enable prediction latent biomechanical variables. methodology yields reactive strategy, which evaluate assisted walking task with robotic lower limb prosthesis. Building upon learned primitives, also present model-predictive (MPC) that actively steers toward movement regimes. We compare introduced strategies highlight framework's ability generate ergonomic, biomechanically assistive prosthetic control. A rich analysis constrained MPC shows 20× reduction large perturbations system. empirically demonstrate 16% vertical reaction forces real-world jumping experiments utilizing our examine other optimal simulated experiments.