Generation of Human-like Motion for Humanoid Robots Based on Marker-based Motion Capture Data

Abstract

To increase acceptance for humanoids in everyday situations, it is essential that motions of humanoid robots become more human-like. A proper approach to achieve this requirement is introduced by adopting marker-based human motion capture. In order to efficiently re-use or analyze captured movements on various robots, an intermediate model, named Master Motor Map (MMM), is proposed which decouples representation of motion from its execution on a real robot. Moreover, we present a constrained nonlinear optimization to adapt pre-captured motions to our robot Armar-III preserving necessary motion characteristics and preventing the robot from approaching specific limitations.