Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2410.20786 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909367893229568 |
|---|---|
| author | Ma, Jianmina Ji, Jingtian Gao, Yue |
| author_facet | Ma, Jianmina Ji, Jingtian Gao, Yue |
| contents | Constrained reinforcement learning has achieved promising progress in safety-critical fields where both rewards and constraints are considered. However, constrained reinforcement learning methods face challenges in striking the right balance between task performance and constraint satisfaction and it is prone for them to get stuck in over-conservative or constraint violating local minima. In this paper, we propose Adversarial Constrained Policy Optimization (ACPO), which enables simultaneous optimization of reward and the adaptation of cost budgets during training. Our approach divides original constrained problem into two adversarial stages that are solved alternately, and the policy update performance of our algorithm can be theoretically guaranteed. We validate our method through experiments conducted on Safety Gymnasium and quadruped locomotion tasks. Results demonstrate that our algorithm achieves better performances compared to commonly used baselines. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_20786 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Adversarial Constrained Policy Optimization: Improving Constrained Reinforcement Learning by Adapting Budgets Ma, Jianmina Ji, Jingtian Gao, Yue Machine Learning Robotics 68T01 I.2.6 Constrained reinforcement learning has achieved promising progress in safety-critical fields where both rewards and constraints are considered. However, constrained reinforcement learning methods face challenges in striking the right balance between task performance and constraint satisfaction and it is prone for them to get stuck in over-conservative or constraint violating local minima. In this paper, we propose Adversarial Constrained Policy Optimization (ACPO), which enables simultaneous optimization of reward and the adaptation of cost budgets during training. Our approach divides original constrained problem into two adversarial stages that are solved alternately, and the policy update performance of our algorithm can be theoretically guaranteed. We validate our method through experiments conducted on Safety Gymnasium and quadruped locomotion tasks. Results demonstrate that our algorithm achieves better performances compared to commonly used baselines. |
| title | Adversarial Constrained Policy Optimization: Improving Constrained Reinforcement Learning by Adapting Budgets |
| topic | Machine Learning Robotics 68T01 I.2.6 |
| url | https://arxiv.org/abs/2410.20786 |