DiffStitch: Boosting Offline Reinforcement Learning with Diffusion-based Trajectory Stitching
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866910339251044352 |
|---|---|
| author | Li, Guanghe Shan, Yixiang Zhu, Zhengbang Long, Ting Zhang, Weinan |
| author_facet | Li, Guanghe Shan, Yixiang Zhu, Zhengbang Long, Ting Zhang, Weinan |
| contents | In offline reinforcement learning (RL), the performance of the learned policy highly depends on the quality of offline datasets. However, in many cases, the offline dataset contains very limited optimal trajectories, which poses a challenge for offline RL algorithms as agents must acquire the ability to transit to high-reward regions. To address this issue, we introduce Diffusion-based Trajectory Stitching (DiffStitch), a novel diffusion-based data augmentation pipeline that systematically generates stitching transitions between trajectories. DiffStitch effectively connects low-reward trajectories with high-reward trajectories, forming globally optimal trajectories to address the challenges faced by offline RL algorithms. Empirical experiments conducted on D4RL datasets demonstrate the effectiveness of DiffStitch across RL methodologies. Notably, DiffStitch demonstrates substantial enhancements in the performance of one-step methods (IQL), imitation learning methods (TD3+BC), and trajectory optimization methods (DT). |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2402_02439 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | DiffStitch: Boosting Offline Reinforcement Learning with Diffusion-based Trajectory Stitching Li, Guanghe Shan, Yixiang Zhu, Zhengbang Long, Ting Zhang, Weinan Machine Learning Artificial Intelligence In offline reinforcement learning (RL), the performance of the learned policy highly depends on the quality of offline datasets. However, in many cases, the offline dataset contains very limited optimal trajectories, which poses a challenge for offline RL algorithms as agents must acquire the ability to transit to high-reward regions. To address this issue, we introduce Diffusion-based Trajectory Stitching (DiffStitch), a novel diffusion-based data augmentation pipeline that systematically generates stitching transitions between trajectories. DiffStitch effectively connects low-reward trajectories with high-reward trajectories, forming globally optimal trajectories to address the challenges faced by offline RL algorithms. Empirical experiments conducted on D4RL datasets demonstrate the effectiveness of DiffStitch across RL methodologies. Notably, DiffStitch demonstrates substantial enhancements in the performance of one-step methods (IQL), imitation learning methods (TD3+BC), and trajectory optimization methods (DT). |
| title | DiffStitch: Boosting Offline Reinforcement Learning with Diffusion-based Trajectory Stitching |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2402.02439 |