A two-level model of anticipation-based motor learning for whole body motion

Abstract

We present a model of motor learning based on a combination of Operational Space Control and Optimal Control. Anticipatory processes are used both in the learning of the dynamics model of the system and in the coordination between both types of control. In order to illustrate the proposed model and associated control method, we apply these principles to the control of a simplified virtual humanoid performing a stand-up task starting from a crouching posture. © 2009 Springer Berlin Heidelberg.