SafeDiSTL: Decomposing Joint Constraints for Safe Multi-Agent Reinforcement Learning
[Uczenie maszynowe] Autorzy: Hongrui Zheng, Aditya Singh, Rahul Mangharam
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 t…
- Kategoria arxiv
- cs.LG
- Kategoria
- Uczenie maszynowe