Safe Reinforcement Learning in Black-Box Environments via Adaptive Shielding

2024/05/28 by Daniel Bethell, Simos Gerasimou, Bethell, Daniel +5 · 1 citation
Engineering · #Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Smart Grid Security and Resilience

paper · doi:10.48550/arxiv.2405.18180

openalex publication_date 2024/05/28 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/29

Abstract

Empowering safe exploration of reinforcement learning (RL) agents during training is a critical challenge towards their deployment in many real-world scenarios. When prior knowledge of the domain or task is unavailable, training RL agents in unknown, black-box environments presents an even greater safety risk. We introduce ADVICE (Adaptive Shielding with a Contrastive Autoencoder), a novel post-shielding technique that distinguishes safe and unsafe features of state-action pairs during training, and uses this knowledge to protect the RL agent from executing actions that yield likely hazardous outcomes. Our comprehensive experimental evaluation against state-of-the-art safe RL exploration techniques shows that ADVICE significantly reduces safety violations (approx 50%) during training, with a competitive outcome reward compared to other techniques.

Cited by

• Safe But Not Sorry: Reducing Over-Conservatism in Safety Critics via Uncertainty-Aware Modulation

Related

• Safety through Permissibility: Shield Construction for Fast and Safe Reinforcement Learning
• Safe Reinforcement Learning via Shielding under Partial Observability
• Dynamic Shielding for Reinforcement Learning in Black-Box Environments
• Safe Multi-Agent Reinforcement Learning via Shielding
• Safe and Robust Reinforcement Learning: Principles and Practice