Design And Implementation Of A Multi-group Intention Recognition System For Wargame

With the rapid development of computer technology,wargame with the support of computers have shown a development trend of complexity and reality.In a more complex and real wargame,the target information that the wargame commander needs to consider is more massive and more detailed.This makes it more difficult for wargame commanders to extract higher-level,abstract game situation information from these target informations.Therefore,the situation assessment technology that simplifies target information and extracts key information is a focus of research in current wargames.This paper studies the target clustering and intention recognition techniques commonly used in situation assessment.In this paper,target clustering and intention recognition are combined,and the intention recognition technology for a single target entity is converted into a technology for intention recognition for a target group.And through target clustering into multiple target groups,and the identification of group intentions,the process of situation estimation is accelerated.For target clustering,in addition to simplifying situational information,this paper additionally introduces the need to provide support for intent recognition.Aiming at the problem that the target unit data that needs to be clustered is large in scale,high in dimension and inconsistent in dimensions,this paper adopts the clustering algorithm combining autoencoder and kmeans to solve the problem,first from supplementing zero value to fixed length,supplementing additional features,etc.The target data is processed,and then the dimensionality reduction of the target unit data is performed by the autoencoder to obtain the salient features of the data.Finally,the kmeans algorithm is used to cluster the salient features of these target units,and the target units are aggregated into one target.The unit group not only simplifies the situation information,but also provides the initial conditions for the subsequent group intention identification.For group intention recognition,this paper uses RNN to identify the intentions of several key targets in the group as group intentions,completes the intention recognition of target units in a group,and accelerates the overall intention recognition process.Aiming at the problem that the lowpurity group under this method will affect the accuracy of intent recognition,this paper proposes a method to improve the purity of the group based on subdivision.This method judges whether the current group is suitable for group intention recognition by whether several key target intentions identified by RNN are consistent.The purity of the group and ultimately improve the accuracy of group intention recognition.In addition,in view of the difficulty of combining the current wargame simulation platform with intention recognition algorithm research,this paper designs and implements a wargame-oriented intention recognition simulation system.The system supports users to configure a wargame environment that meets their own needs,and generates simulation data for users to train algorithm models through functions such as simulation deduction,playback,and replay.This method supports the embedding of the user’s own algorithm model into the wargame simulation process.Finally,the above functions are used to provide users with a simulation platform that is convenient for studying the problem of target grouping and intention recognition in wargames.