Low Dimensional Motor Skill Learning Using Coactivation
We propose an approach for motor skill learning in highly articulated characters by exploring low-dimensional joint coactivation spaces. Analysis of human motion reveals that many tasks have lower dimensionality than the character's full degrees of freedom (DOFs), with joints often moving in synchrony.
Highlights Video
Conference Presentation
Overview
We exploit the redundancy in human motion by extracting task-specific joint coactivations from human motion data, capturing synchronized patterns that reduce the control space. Using deep reinforcement learning, we train controllers to excite these coactivations, enabling humanlike control with only a few dimensions.
Key Results
- 10-dimensional controller for a 28 DOF character
- High-fidelity locomotion even with sparse rewards
- Demonstrated on various motor tasks
- Flexibility with different reward functions
Publication
A. Ranganath, P. Xu, I. Karamouzas, and V. Zordan, "Low Dimensional Motor Skill Learning Using Coactivation", in Motion, Interaction and Games, 2019, Newcastle, UK, Oct 2019.
Download PDF