arXiv cs.AI· Hongrui Zheng, Aditya Singh, Rahul Mangharam·· 6 天前AI 评分23
SafeDiSTL：为安全多智能体强化学习分解联合约束
SafeDiSTL: Decomposing Joint Constraints for Safe Multi-Agent Reinforcement Learning
AI 导读
SafeDiSTL 是一个约束多智能体强化学习框架，利用 Signal Temporal Logic（STL）鲁棒性度量回合级违规，并通过可微动力学得到的动作敏感度将联合约束成本增量分配到各智能体，实现联合约束下的信用分配。在协作载荷运输、联网导航和网络化交通控制任务中，其安全成功率分别达到约束基线的 1.6-5.1 倍、提升 11.1%，交通吞吐量提升 50-68%。
正文 · 原文
Abstract:Cooperative multi-agent tasks often impose constraints on a shared objective or collective safety. A team level violation cost function can measure this constraint outcome, but does not distinguish which agents' actions most strongly influence that violation. We present SafeDiSTL, a constrained multi-agent reinforcement learning framework that allows for credit assignment even with joint constraint measurements. We use Signal Temporal Logic (STL) robustness to measure episode-level violation through running-maximum cost increments. Smooth full-episode robustness and differentiable dynamics provide action sensitivities, which allocate each cost increment across agents at the same decision time. These sensitivities account for delayed effects when weighting agents' shares of each cost increment. We study the framework in cooperative payload transport, connected navigation, and networked traffic control tasks. SafeDiSTL achieves 1.6-5.1x the safe-success rate in transport, improves navigation safe-success by 11.1%, and delivers 50-68% more throughput in traffic when compared to constrained baselines. Results demonstrate a strong improvement of the learned policy in dealing with shared constraint while retaining the performance of unconstrained policies.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.13737 [cs.LG] |
| (or arXiv:2610.13737v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.13737 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Hongrui Zheng [view email]
[v1]
Fri, 9 Oct 2026 15:32:06 UTC (284 KB)
来源：arXiv cs.AI · arxiv.org