Literature¶

The repository includes paper PDFs under docs/Paper References. These papers motivate the safety and assume-guarantee machinery, but the implementation is documented in the user pages and docs/AI.

Included References¶

• Compositional shielding for safe reinforcement learning in partial observability
• Compositional Shielding and Reinforcement Learning for Multi-Agent Systems
• Compositional Circular Assume-Guarantee Rules Cannot Be Sound and Complete
• Automated Assume-Guarantee Reasoning by Abstraction Refinement
• Assume-Guarantee Reasoning with Local Specifications
• Assume-Guarantee Distributed Synthesis

How They Map To This Repository¶

The repository combines three ideas:

• LTL safety monitoring and shield synthesis,
• local specifications for agents,
• contract-style assume-guarantee certification for multi-agent learning.

The current implementation uses centralized certification for a candidate contract profile, then decentralized execution with per-agent action masks.