Research And Development Of Intelligent Combat Behavior Modeling And Simulation System

With the development of simulation technology,combat simulation systems have been widely used to conduct command strategy training for military personnel,evaLuate and optimize established combat plans,and gradually become an important technical means for studying military combat issues.The modern combat system is a typical hybrid game system,and its modeling should fully consider the hybrid characteristics contained in it.However,the existing combat simulation system cannot fully and effectively describe the hybrid characteristics of the aforementioned combat system,so explore the corresponding modeling Methods and behavior modeling of combat entities,and the development of intelligent combat simulation systems,is a topic that needs urgent research.In response to the above-mentioned needs and current technical problems,this article carries out the following work.(1)Aiming at the insufficient description of various hybrid characteristics such as discrete,continuous,random,time delay and decision-making in modern combat systems,in-depth investigation of domestic and foreign combat system modeling methods,behavior modeling and combat simulation systems,and summarize current modeling The method is insufficient,and the combat system architecture of "behavior layer-event layer-physical layer" based on Hybrid Stochastic Timed Petri Net(HSTPN)and the corresponding combat system modeling method are proposed,and the United States is initiated to Libya The "El Dorado Canyon"operation was modeled to verify the effectiveness of the modeling method.(2)Aiming at the shortcomings of the large information integrity requirements,low efficiency in facing complex problems and poor model versatility in the action rule modeling method.A K-means clustering method based on 8 types of attributes such as the attribution party,target and range of the combat entity is proposed,and the combat entity model is fuzzified with different granularities,which is convenient for military personnel to conduct combat systems on different scales of campaigns.Modeling.For the aggregated entities after clustering,a decision tree-based action rule model design method is proposed,and the possibility of using the decision tree method is analyzed,which provides a specific solution for the formatted expression and formal description of the action rule model.Taking the long-range air defense missile in the "El Dorado Canyon" operation as an example,the action rule model was realized and realized by programming in Lua script.(3)Aiming at the characteristics of limited combat data and insufficient model generalization ability in command and decision behavior modeling.Deep reinforcement learning Dueling DDQN algorithm is used to build a command decision model,the command decision process is regarded as a Markov decision process,and the design method and training process of the command decision model are given.Take the "El Dorado Canyon" operation as an example to build a US military command Decision model.Designing a deep learning engagement decision model based on battle history data.It can be used for battlefield situation prediction and battle results judgment in the course of battle,and can also replace the battle simulation environment for pre-training of command and decision-making models and improve the training of command and decision-making models.effectiveness.(4)Achieve a combat simulation system based on the "behavior layer-event layer-physical layer" architecture,and design a human-computer interaction interface,use DirectX technology to realize the visual deduction of the combat process,and be able to visually observe the combat process and simulation results.The "El Dorado Canyon" operation was used as a case to complete the application verification of the combat system.