Motion Transformation with Dynamic Consistency Using Nonlinear State-Space Mapping

Abstract

This paper proposes a motion transformation method using nonlinear state-space mapping. A human motion is measured by a motion capture system and it is transformed to a “realizable” robot motion. “Realizable” requires existence of an input sequence that satisfies the dynamical consistency of the robot dynamics. The nonlinear state-space mapping is represented by a polynomial of power of state variable and its parameters are obtained so that the constrains of velocity and acceleration are satisfied, and the human motion is deformed simultaneously. Moreover, the transformed motion is decided by setting the reference kinetic energy. The proposed method is applied to a tap dancing robot and a robot motion is generated by an autonomous system based on orbit attractor.