基于Stackelberg策略的多Agent强化学习警力巡逻路径规划

发布时间：2018-02-16 13:10

  本文关键词： 巡逻路线规划 Stackelberg强均衡策略 多agent 强化学习　出处：《北京理工大学学报》2017年01期 　论文类型：期刊论文

【摘要】：为解决现有的巡逻路径规划算法仅仅能够处理双人博弈和忽略攻击者存在的问题,提出一种新的基于多agent的强化学习算法.在给定攻击目标分布的情况下,规划任意多防御者和攻击者条件下的最优巡逻路径.考虑到防御者与攻击者选择策略的非同时性,采用了Stackelberg强均衡策略作为每个agent选择策略的依据.为了验证算法,在多个巡逻任务中进行了测试.定量和定性的实验结果证明了算法的收敛性和有效性.
[Abstract]:In order to solve the problem that the existing patrol path planning algorithms can only deal with the two-game and ignore the attackers, a new reinforcement learning algorithm based on multiple agent is proposed. The optimal patrol path is planned under the condition of arbitrary multiple defenders and attackers. Considering the non-synchronization of the choice strategy between the defender and the attacker, the Stackelberg strong equilibrium strategy is adopted as the basis of each agent selection strategy. The results of quantitative and qualitative experiments show that the algorithm is convergent and effective.
【作者单位】： 中国人民公安大学网络安全保卫学院;
【基金】：中国人民公安大学基本科研业务费项目(2014JKF01132)
【分类号】：D631.1;TP18
，

本文编号：1515593