Almost surely safe exploration and exploitation for deep reinforcement learning with state safety estimation

DOI10.1016/J.INS.2024.120261MaRDI QIDQ6495127

Yan-Jie Li, Qi Liu, Ke Lin, Shiyu Chen, Duantengchuan Li, Xiongtao Shi

Publication date: 30 April 2024

Published in: Information Sciences (Search for Journal in Brave)

zbMATH Keywords

Gaussian processreinforcement learningrisk-sensitive reinforcement learningsafe explorationstate safety estimation

Mathematics Subject Classification ID

Computer science (68-XX) Game theory, economics, finance, and other social and behavioral sciences (91-XX)

Cites Work

• Unnamed Item
• Unnamed Item
• Unnamed Item
• Provably safe and robust learning-based model predictive control
• Simple statistical gradient-following algorithms for connectionist reinforcement learning
• Constrained model predictive control: Stability and optimality
• How Deep Are Deep Gaussian Processes?
• Risk-Constrained Reinforcement Learning with Percentile Risk Criteria
• A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play
• A General Safety Framework for Learning-Based Control in Uncertain Robotic Systems

This page was built for publication: Almost surely safe exploration and exploitation for deep reinforcement learning with state safety estimation