Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2601.18142 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912848993583104 |
|---|---|
| author | Zhang, Mingxu Zhang, Huicheng Ji, Jiaming Yang, Yaodong Sun, Ying |
| author_facet | Zhang, Mingxu Zhang, Huicheng Ji, Jiaming Yang, Yaodong Sun, Ying |
| contents | Safe reinforcement learning (Safe RL) seeks to maximize rewards while satisfying safety constraints, typically addressed through Lagrangian-based methods. However, existing approaches, including PID and classical Lagrangian methods, suffer from oscillations and frequent safety violations due to parameter sensitivity and inherent phase lag. To address these limitations, we propose ADRC-Lagrangian methods that leverage Active Disturbance Rejection Control (ADRC) for enhanced robustness and reduced oscillations. Our unified framework encompasses classical and PID Lagrangian methods as special cases while significantly improving safety performance. Extensive experiments demonstrate that our approach reduces safety violations by up to 74%, constraint violation magnitudes by 89%, and average costs by 67\%, establishing superior effectiveness for Safe RL in complex environments. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_18142 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Enhance the Safety in Reinforcement Learning by ADRC Lagrangian Methods Zhang, Mingxu Zhang, Huicheng Ji, Jiaming Yang, Yaodong Sun, Ying Machine Learning Safe reinforcement learning (Safe RL) seeks to maximize rewards while satisfying safety constraints, typically addressed through Lagrangian-based methods. However, existing approaches, including PID and classical Lagrangian methods, suffer from oscillations and frequent safety violations due to parameter sensitivity and inherent phase lag. To address these limitations, we propose ADRC-Lagrangian methods that leverage Active Disturbance Rejection Control (ADRC) for enhanced robustness and reduced oscillations. Our unified framework encompasses classical and PID Lagrangian methods as special cases while significantly improving safety performance. Extensive experiments demonstrate that our approach reduces safety violations by up to 74%, constraint violation magnitudes by 89%, and average costs by 67\%, establishing superior effectiveness for Safe RL in complex environments. |
| title | Enhance the Safety in Reinforcement Learning by ADRC Lagrangian Methods |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2601.18142 |