Safe Reinforcement Learning using Finite-Horizon Gradient-based Estimation

Juntao Dai, Yaodong Yang, Qian Zheng, Gang Pan
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:9872-9903, 2024.

Abstract

A key aspect of Safe Reinforcement Learning (Safe RL) involves estimating the constraint condition for the next policy, which is crucial for guiding the optimization of safe policy updates. However, the existing Advantage-based Estimation (ABE) method relies on the infinite-horizon discounted advantage function. This dependence leads to catastrophic errors in finite-horizon scenarios with non-discounted constraints, resulting in safety-violation updates. In response, we propose the first estimation method for finite-horizon non-discounted constraints in deep Safe RL, termed Gradient-based Estimation (GBE), which relies on the analytic gradient derived along trajectories. Our theoretical and empirical analyses demonstrate that GBE can effectively estimate constraint changes over a finite horizon. Constructing a surrogate optimization problem with GBE, we developed a novel Safe RL algorithm called Constrained Gradient-based Policy Optimization (CGPO). CGPO identifies feasible optimal policies by iteratively resolving sub-problems within trust regions. Our empirical results reveal that CGPO, unlike baseline algorithms, successfully estimates the constraint functions of subsequent policies, thereby ensuring the efficiency and feasibility of each update.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-dai24d, title = {Safe Reinforcement Learning using Finite-Horizon Gradient-based Estimation}, author = {Dai, Juntao and Yang, Yaodong and Zheng, Qian and Pan, Gang}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {9872--9903}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/dai24d/dai24d.pdf}, url = {https://proceedings.mlr.press/v235/dai24d.html}, abstract = {A key aspect of Safe Reinforcement Learning (Safe RL) involves estimating the constraint condition for the next policy, which is crucial for guiding the optimization of safe policy updates. However, the existing Advantage-based Estimation (ABE) method relies on the infinite-horizon discounted advantage function. This dependence leads to catastrophic errors in finite-horizon scenarios with non-discounted constraints, resulting in safety-violation updates. In response, we propose the first estimation method for finite-horizon non-discounted constraints in deep Safe RL, termed Gradient-based Estimation (GBE), which relies on the analytic gradient derived along trajectories. Our theoretical and empirical analyses demonstrate that GBE can effectively estimate constraint changes over a finite horizon. Constructing a surrogate optimization problem with GBE, we developed a novel Safe RL algorithm called Constrained Gradient-based Policy Optimization (CGPO). CGPO identifies feasible optimal policies by iteratively resolving sub-problems within trust regions. Our empirical results reveal that CGPO, unlike baseline algorithms, successfully estimates the constraint functions of subsequent policies, thereby ensuring the efficiency and feasibility of each update.} }
Endnote
%0 Conference Paper %T Safe Reinforcement Learning using Finite-Horizon Gradient-based Estimation %A Juntao Dai %A Yaodong Yang %A Qian Zheng %A Gang Pan %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-dai24d %I PMLR %P 9872--9903 %U https://proceedings.mlr.press/v235/dai24d.html %V 235 %X A key aspect of Safe Reinforcement Learning (Safe RL) involves estimating the constraint condition for the next policy, which is crucial for guiding the optimization of safe policy updates. However, the existing Advantage-based Estimation (ABE) method relies on the infinite-horizon discounted advantage function. This dependence leads to catastrophic errors in finite-horizon scenarios with non-discounted constraints, resulting in safety-violation updates. In response, we propose the first estimation method for finite-horizon non-discounted constraints in deep Safe RL, termed Gradient-based Estimation (GBE), which relies on the analytic gradient derived along trajectories. Our theoretical and empirical analyses demonstrate that GBE can effectively estimate constraint changes over a finite horizon. Constructing a surrogate optimization problem with GBE, we developed a novel Safe RL algorithm called Constrained Gradient-based Policy Optimization (CGPO). CGPO identifies feasible optimal policies by iteratively resolving sub-problems within trust regions. Our empirical results reveal that CGPO, unlike baseline algorithms, successfully estimates the constraint functions of subsequent policies, thereby ensuring the efficiency and feasibility of each update.
APA
Dai, J., Yang, Y., Zheng, Q. & Pan, G.. (2024). Safe Reinforcement Learning using Finite-Horizon Gradient-based Estimation. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:9872-9903 Available from https://proceedings.mlr.press/v235/dai24d.html.

Related Material