知识与数据互补的战术级兵棋行为决策框架设计与实现  

Framework Design and Application for Tactical-level Wargame Behavior Decision-making Based on Complementary Knowledge and Data

作　　者：刘满 张宏军 程恺 郝文宁 王之腾 LIU Man;ZHANG Hongjun;CHENG Kai;HAO Wenning;WANG Zhiteng(Unit 73131 of PLA,Zhangzhou Fujian 363000,China;College of Command and Control Engineering,Army Engineering University of PLA,Nanjing Jiangsu 210007,China;College of Communication Engineering,Army Engineering University of PLA,Nanjing Jiangsu 210007,China)

机构地区：[1]解放军73131部队,福建漳州363000 [2]陆军工程大学指挥控制工程学院,江苏南京210007 [3]陆军工程大学通信工程学院,江苏南京210007

出　　处：《指挥与控制学报》2023年第2期182-191,共10页Journal of Command and Control

基　　金：国家自然科学基金(61806221)资助。

摘　　要：战术级兵棋以随机的方式模拟战争中的动态过程,能够为军事智能决策技术提供贴近真实战争的决策背景和试验环境.提出了知识与数据互补的行为决策框架,用于兵棋多实体的指挥控制.该框架一定程度上解决了传统基于知识推理决策中行为模式固定、迁移能力不强的缺点,也解决了基于兵棋数据挖掘的软决策算法对大量人类高质量复盘数据的需求,将知识、数据与学习的方法综合起来,形成基于知识推理的决策算法处理宏观动作,基于数据挖掘的软决策算法处理微观动作,通过自对抗复盘数据进行迭代学习,提升决策模型能力.基于该框架,设计并实现了一个兵棋人工智能(artificial intelligence,AI),该兵棋AI在全国性智能兵棋比赛中取得较好成绩,并体现出灵活性高、泛化性好的特性.The tactical-level wargame simulates the dynamic process of war in a random way,which can provide a decision-making background and test environment closer to real war for military intelligent decision-making technology.A behavior decision-making framework with complementary knowledge and data is proposed for the command and control of wargame multi-agent.The framework solves the shortcomings of fixed behavior pattern and weak migration ability of traditional decision-making based on knowledge reasoning to a certain extent,and also solves the requirement of data mining algorithm for a large number of human high-quality replays of the soft decision-making algorithm based on wargame data mining.It integrates knowledge,data and learning method to form a decision-making algorithm,the macro actions are processed by the knowledge reasoning based decision-making algorithm and the mirco actions are processed by the data mining-based soft decision-making algorithm.The ability of decision-making model is improved by iterative learning with self-confrontation replay data.Based on this framework,a wargame AI is designed and realized,which has achieved good results in a national intelligent wargame competition,and reflects the characteristics of high flexibility and good generalization.

关 键 词：战术级兵棋 行为决策 知识推理 数据挖掘 迭代学习 

分 类 号：E91[军事] TP18[自动化与计算机技术—控制理论与控制工程]