Enabling Stateful Behaviors for Diffusion-based Policy Learning
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866916332417581056 |
|---|---|
| author | Liu, Xiao Weigend, Fabian Zhou, Yifan Amor, Heni Ben |
| author_facet | Liu, Xiao Weigend, Fabian Zhou, Yifan Amor, Heni Ben |
| contents | While imitation learning provides a simple and effective framework for policy learning, acquiring consistent actions during robot execution remains a challenging task. Existing approaches primarily focus on either modifying the action representation at data curation stage or altering the model itself, both of which do not fully address the scalability of consistent action generation. To overcome this limitation, we introduce the Diff-Control policy, which utilizes a diffusion-based model to learn the action representation from a state-space modeling viewpoint. We demonstrate that we can reduce diffusion-based policies' uncertainty by making it stateful through a Bayesian formulation facilitated by ControlNet, leading to improved robustness and success rates. Our experimental results demonstrate the significance of incorporating action statefulness in policy learning, where Diff-Control shows improved performance across various tasks. Specifically, Diff-Control achieves an average success rate of 72% and 84% on stateful and dynamic tasks, respectively. Project page: https://github.com/ir-lab/Diff-Control |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2404_12539 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Enabling Stateful Behaviors for Diffusion-based Policy Learning Liu, Xiao Weigend, Fabian Zhou, Yifan Amor, Heni Ben Robotics While imitation learning provides a simple and effective framework for policy learning, acquiring consistent actions during robot execution remains a challenging task. Existing approaches primarily focus on either modifying the action representation at data curation stage or altering the model itself, both of which do not fully address the scalability of consistent action generation. To overcome this limitation, we introduce the Diff-Control policy, which utilizes a diffusion-based model to learn the action representation from a state-space modeling viewpoint. We demonstrate that we can reduce diffusion-based policies' uncertainty by making it stateful through a Bayesian formulation facilitated by ControlNet, leading to improved robustness and success rates. Our experimental results demonstrate the significance of incorporating action statefulness in policy learning, where Diff-Control shows improved performance across various tasks. Specifically, Diff-Control achieves an average success rate of 72% and 84% on stateful and dynamic tasks, respectively. Project page: https://github.com/ir-lab/Diff-Control |
| title | Enabling Stateful Behaviors for Diffusion-based Policy Learning |
| topic | Robotics |
| url | https://arxiv.org/abs/2404.12539 |