基于知识图谱的仿真想定智能生成方法  

Intelligent Generation Method of Simulation Scenarios Based on Knowledge Graph

作　　者：安靖 司光亚[3] 周杰[2] 韩旭 AN Jing;SI Guangya;ZHOU Jie;HAN Xu(Graduate School,National Defense University,Beijing 100091,China;Joint Logistics College,National Defense University,Beijing 100858,China;Joint Operation College,National Defense University,Beijing 100091,China)

机构地区：[1]国防大学研究生院,北京100091 [2]国防大学联合勤务学院,北京100858 [3]国防大学联合作战学院,北京100091

出　　处：《指挥与控制学报》2023年第1期103-109,共7页Journal of Command and Control

基　　金：全军军事类研究生资助课题(JY2020B031)资助。

摘　　要：仿真想定的开发,存在着专业性强、定制要求高、迁移性差等问题.针对这一系列问题,设计并实现一种基于知识图谱的仿真想定智能生成方法.通过基于强化学习的全局特征链接多源仿真模型实体,构建了知识图谱;基于该知识图谱,设计了适用于仿真模型匹配的实体对齐算法和基于文档对象模型(document object model,DOM)的脚本生成方法,实现仿真想定的智能映射,以“立体投送”行动为实例展开实验,实验结果表明:生成的仿真想定能够驱动仿真系统按照预定行动时序运行,方法有效.The development of simulation scenarios has such problems,as strong professionalism,high customization requirements and poor mobility,etc.Aiming at these problems,an intelligent generation method of simulation scenarios based on knowledge graph(KG)is designed and implemented.Firstly,the KG is constructed by linking multi-source simulation model entities with global features based on reinforcement learning(RL).Based on this,an entity alignment algorithm suitable for simulation model matching and a script generation method based on document object model(DOM)are designed to realize intelligent mapping of simulation scenarios.Finally,the"multi-dimension projection"action is taken as an example,the experimental results show that the generated simulation scenarios can drive the simulation system to run according to the scheduled action sequence,and the method is effective.

关 键 词：仿真想定 知识图谱 强化学习 实体对齐 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术] E91[自动化与计算机技术—计算机科学与技术]