Adapting attackers and defenders patrolling strategies: A reinforcement learning approach for Stackelberg security games
作者：
Highlights：
• We consider a Reinforcement Learning process for Stackelberg security games.
• The architecture involves the Adaptive Primary Learning and the Actor–critic modules.
• We provide a game-theoretic formulation method based on Markov decision process.
• We design the extraproximal method to find the Strong Lp-Stackelberg/Nash equilibrium.
• We provide an efficient algorithm that accelerate the reinforcement learning process.
摘要
•We consider a Reinforcement Learning process for Stackelberg security games.•The architecture involves the Adaptive Primary Learning and the Actor–critic modules.•We provide a game-theoretic formulation method based on Markov decision process.•We design the extraproximal method to find the Strong Lp-Stackelberg/Nash equilibrium.•We provide an efficient algorithm that accelerate the reinforcement learning process.
论文关键词：Security games,Reinforcement learning,Stackelberg games,Behavioral games,Multiple players,Strong Stackelberg/Nash equilibrium
论文评审过程：Received 8 February 2017, Revised 11 October 2017, Accepted 26 December 2017, Available online 9 January 2018, Version of Record 30 April 2018.
论文官网地址：https://doi.org/10.1016/j.jcss.2017.12.004
