SafeDiSTL: Decomposing Joint Constraints for Safe Multi-Agent Reinforcement Learning
We present SafeDiSTL, a constrained multi-agent reinforcement learning framework that allows for credit assignment even with joint constraint measurements.
ProofPaper ↗
Key points
- 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 use Signal Temporal Logic (STL) robustness to measure episode-level violation through running-maximum cost increments.
- We study the framework in cooperative payload transport, connected navigation, and networked traffic control tasks.
Sources (1)
- [1]SafeDiSTL: Decomposing Joint Constraints for Safe Multi-Agent Reinforcement LearningarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 03:32 PM
We present SafeDiSTL, a constrained multi-agent reinforcement learning framework that allows for credit assignment even with joint constraint measurements.
Cooperative multi-agent tasks often impose constraints on a shared objective or collective safety.
Extractive summary: sentences quoted from the sources.
