VMP: Versatile Motion Priors for Robustly Tracking Motion on Physical Characters
| dc.contributor.author | Serifi, Agon | en_US |
| dc.contributor.author | Grandia, Ruben | en_US |
| dc.contributor.author | Knoop, Espen | en_US |
| dc.contributor.author | Gross, Markus | en_US |
| dc.contributor.author | Bächer, Moritz | en_US |
| dc.contributor.editor | Skouras, Melina | en_US |
| dc.contributor.editor | Wang, He | en_US |
| dc.date.accessioned | 2024-08-20T08:42:54Z | |
| dc.date.available | 2024-08-20T08:42:54Z | |
| dc.date.issued | 2024 | |
| dc.description.abstract | Recent progress in physics-based character control has made it possible to learn policies from unstructured motion data. However, it remains challenging to train a single control policy that works with diverse and unseen motions, and can be deployed to real-world physical robots. In this paper, we propose a two-stage technique that enables the control of a character with a full-body kinematic motion reference, with a focus on imitation accuracy. In a first stage, we extract a latent space encoding by training a variational autoencoder, taking short windows of motion from unstructured data as input. We then use the embedding from the time-varying latent code to train a conditional policy in a second stage, providing a mapping from kinematic input to dynamics-aware output. By keeping the two stages separate, we benefit from self-supervised methods to get better latent codes and explicit imitation rewards to avoid mode collapse. We demonstrate the efficiency and robustness of our method in simulation, with unseen user-specified motions, and on a bipedal robot, where we bring dynamic motions to the real world. | en_US |
| dc.description.number | 8 | |
| dc.description.sectionheaders | Character Animation II: Control | |
| dc.description.seriesinformation | Computer Graphics Forum | |
| dc.description.volume | 43 | |
| dc.identifier.doi | 10.1111/cgf.15175 | |
| dc.identifier.issn | 1467-8659 | |
| dc.identifier.pages | 11 pages | |
| dc.identifier.uri | https://doi.org/10.1111/cgf.15175 | |
| dc.identifier.uri | https://diglib.eg.org/handle/10.1111/cgf15175 | |
| dc.publisher | The Eurographics Association and John Wiley & Sons Ltd. | en_US |
| dc.subject | CCS Concepts: Computing methodologies → Learning from demonstrations; Learning latent representations; Reinforcement learning; Physical simulation; Animation; Control methods | |
| dc.subject | Computing methodologies → Learning from demonstrations | |
| dc.subject | Learning latent representations | |
| dc.subject | Reinforcement learning | |
| dc.subject | Physical simulation | |
| dc.subject | Animation | |
| dc.subject | Control methods | |
| dc.title | VMP: Versatile Motion Priors for Robustly Tracking Motion on Physical Characters | en_US |