Safe and Generalizable Hierarchical Multi-Agent RL via Constraint Manifold Control

Short summary

Researchers propose a hierarchical multi-agent reinforcement learning framework that enforces hard safety constraints while enabling high-level policy learning. The approach provides theoretical safety guarantees and stable training dynamics, achieving competitive performance with near-perfect safety rates. Empirically demonstrates generalization across varying numbers of agents and obstacles.

• •Hierarchical RL framework bridging control-theoretic safety with learning-based coordination
• •Provides theoretical safety guarantees and stationary learning dynamics
• •Achieves competitive performance with near-perfect safety rates and strong generalization

Generated with AI, which can make mistakes.