基于多智能体博弈强化学习的无人机智能攻击策略生成模型  

UAV intelligent attack strategy generation model based on multi-agent game reinforcement learning

作　　者：赵芷若 曹雷 陈希亮 赖俊 章乐贵 ZHAO Zhiruo;CAO Lei;CHEN Xiliang;LAI Jun;ZHANG Legui(Command and Control Engineering College,Army Engineering University of PLA,Nanjing 210007,China)

机构地区：[1]中国人民解放军陆军工程大学指挥控制工程学院,江苏南京210007

出　　处：《系统工程与电子技术》2023年第10期3165-3171,共7页Systems Engineering and Electronics

摘　　要：如何利用以攻击型无人机(unmanned aerial vehicle,UAV)为代表的新型作战力量增强战斗力,是智能化、无人化战争研究的重点之一。研究了基于多智能体博弈强化学习的无人机智能攻击关键技术,基于马尔可夫随机博弈的基本概念,建立了基于多智能体博弈强化学习的无人机智能攻击策略生成模型,并利用博弈论中“颤抖的手完美”思想提出优化方法,改进了策略模型。仿真实验表明,优化后的算法在原算法基础上有所提升,训练得到的模型可生成多种实时攻击战术,对智能化指挥控制具有较强的现实意义。How to utilize new combat forces represented by offensive unmanned aerial vehicle(UAV)to enhance combat effectiveness is one of the focuses of intelligent and unmanned warfare research.This article is based on the key technology of UAV intelligent attack using multi-agent game reinforcement learning,as well as the basic concept of Markov random games.A model for generating UAV intelligent attack strategies based on multi-agent game reinforcement learning is established,and an optimization method is proposed using the“trembling hand perfect”idea in the game theory to improve the strategy model.Simulation experiments show that the optimized algorithm has improved the original algorithm,and the trained model can generate various real-time attack tactics,which has strong practical significance for intelligent command and control.

关 键 词：多智能体博弈强化学习 马尔可夫随机博弈 无人机 战术策略 

分 类 号：E917[军事]