Offline Trajectory Optimization for Offline Reinforcement Learning
Offline reinforcement learning (RL) aims to learn policies without online explorations. To enlarge the training data, model-based offline RL learns a dynamics model which is utilized as a virtual environment to generate simulation data and enhance policy learning. However, existing data augmentation methods for offline RL suffer from (i) trivial improvement from short-horizon simulation; and (ii) the lack of evaluation and correction for generated data, leading to low-qualified augmentation.
Authors
- Pengjie Ren (ORCID: https://orcid.org/0000-0003-2964-6422)
- Zhumin Chen (ORCID: https://orcid.org/0000-0003-4592-4074)
- Yongli Liang
- Fajie Yuan (ORCID: https://orcid.org/0000-0001-8452-9929)
- Zhaochun Ren (ORCID: https://orcid.org/0000-0002-9076-6565)
- Xin Xin (ORCID: https://orcid.org/0000-0001-6116-9115)
- Ziqi Zhao (ORCID: https://orcid.org/0009-0008-3011-5745)
- Liu Yang (ORCID: https://orcid.org/0009-0007-7508-0964)
- Jun Ma
Institutions
- Leiden University (NL)
- Shandong University (CN)
- Westlake University (CN)
Publication Details
- Published
- 2025-08-03
- DOI
- https://doi.org/10.1145/3711896.3737066
- Primary Topic
- Reinforcement Learning in Robotics
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Natural Science Foundation of China
- Natural Science Foundation of Shandong Province