Probabilistic Safety Guarantees for Markov Decision Processes

• 成果类型: Article
• 署名作者: Wisniewski, Rafal; Bujorianu, Manuela L.
• 署名单位: Aalborg University; University of London; University College London
• 刊物名称: IEEE TRANSACTIONS ON AUTOMATIC CONTROL
• ISSN/ISSBN: 0018-9286
• DOI: 10.1109/TAC.2023.3291952
• 发表日期: 2023
• 页码: 8095-8102
• 关键词: Dynamic programming (DP); Linear programming (LP); Markov decision processes (MDPs); safety

摘要

This article aims to incorporate safety specifications into Markov decision processes. Explicitly, we address the minimization problem up to a stopping time with safety constraints. We establish a formalism leaning upon the evolution equation to achieve our goal. We show how to compute the safety function with dynamic programming. In the last part of this article, we develop several algorithms for safe stochastic optimization using linear and dynamic programming.