New papers propose shielding methods for safe RL

Arxiv CS.LG (machine learning)

Two arXiv papers introduce shielding approaches for safe reinforcement learning. 'Contract-Based Compositional Shielding' addresses multi-agent coordination under global safety constraints. 'Robust Shielding' handles unknown transition dynamics in MDPs.

safe-rlReinforcement learningMulti-agent systems

See 1 more source

Robust Shielding for Safe Reinforcement Learning14 days agoEdwin Hamel-De le Court, Thom Badings, Alessandro Abate, Francesco Belardinelli, Francesco Fabiano

Arxiv CS.LG (machine learning)·Omar AdalatEdwin Hamel-De le CourtFrancesco Belardinelli

1 day ago