Generation of human-like motion on anthropomorphic systems using inverse dynamics

Abstract

This work deals with the generation of human-like whole-body movements on anthropomorphic systems. We propose a general framework to generate robot movements from the definition of ordered stack of tasks and a global resolution scheme that enables to consider different kinds of constraints. We compare qualitatively the robot movements generated from this software with similar recorded human movements. We start with a direct global comparison of body movements. Then we analyze the magnitude of the reconstructed human torques and compare with the simulated robot torques during the motion.