Preference Aligned Diffusion Planner for Quadrupedal Locomotion Control
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866916640393789440 |
|---|---|
| author | Yuan, Xinyi Shang, Zhiwei Wang, Zifan Wang, Chenkai Shan, Zhao Zhu, Meixin Bai, Chenjia Li, Xuelong Wan, Weiwei Harada, Kensuke |
| author_facet | Yuan, Xinyi Shang, Zhiwei Wang, Zifan Wang, Chenkai Shan, Zhao Zhu, Meixin Bai, Chenjia Li, Xuelong Wan, Weiwei Harada, Kensuke |
| contents | Diffusion models demonstrate superior performance in capturing complex distributions from large-scale datasets, providing a promising solution for quadrupedal locomotion control. However, the robustness of the diffusion planner is inherently dependent on the diversity of the pre-collected datasets. To mitigate this issue, we propose a two-stage learning framework to enhance the capability of the diffusion planner under limited dataset (reward-agnostic). Through the offline stage, the diffusion planner learns the joint distribution of state-action sequences from expert datasets without using reward labels. Subsequently, we perform the online interaction in the simulation environment based on the trained offline planner, which significantly diversified the original behavior and thus improves the robustness. Specifically, we propose a novel weak preference labeling method without the ground-truth reward or human preferences. The proposed method exhibits superior stability and velocity tracking accuracy in pacing, trotting, and bounding gait under different speeds and can perform a zero-shot transfer to the real Unitree Go1 robots. The project website for this paper is at https://shangjaven.github.io/preference-aligned-diffusion-legged. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_13586 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Preference Aligned Diffusion Planner for Quadrupedal Locomotion Control Yuan, Xinyi Shang, Zhiwei Wang, Zifan Wang, Chenkai Shan, Zhao Zhu, Meixin Bai, Chenjia Li, Xuelong Wan, Weiwei Harada, Kensuke Robotics Diffusion models demonstrate superior performance in capturing complex distributions from large-scale datasets, providing a promising solution for quadrupedal locomotion control. However, the robustness of the diffusion planner is inherently dependent on the diversity of the pre-collected datasets. To mitigate this issue, we propose a two-stage learning framework to enhance the capability of the diffusion planner under limited dataset (reward-agnostic). Through the offline stage, the diffusion planner learns the joint distribution of state-action sequences from expert datasets without using reward labels. Subsequently, we perform the online interaction in the simulation environment based on the trained offline planner, which significantly diversified the original behavior and thus improves the robustness. Specifically, we propose a novel weak preference labeling method without the ground-truth reward or human preferences. The proposed method exhibits superior stability and velocity tracking accuracy in pacing, trotting, and bounding gait under different speeds and can perform a zero-shot transfer to the real Unitree Go1 robots. The project website for this paper is at https://shangjaven.github.io/preference-aligned-diffusion-legged. |
| title | Preference Aligned Diffusion Planner for Quadrupedal Locomotion Control |
| topic | Robotics |
| url | https://arxiv.org/abs/2410.13586 |