Research papers · 2025.01
SMARLA: A Safety Monitoring Approach for Deep Reinforcement Learning Agents
SMARLA: A Safety Monitoring Approach for Deep Reinforcement Learning Agents
Overview
SMARLA predicts reinforcement-learning safety violations from Q-values and abstracted execution states. Three case studies evaluate this black-box monitor, while confidence-based decision criteria explore the trade-off between early warning and false alarms.
Author-supplied source information
Author-supplied information on arXiv.
in IEEE Transactions on Software Engineering, vol. 51, no. 01, pp. 82-105, Jan. 2025
Version: arXiv v4 — manuscript revision