A wargame situation prediction method based on a graph neural network GraphVAE

Abstract

In the process of the battle between the enemy and us on the battlefield, accurately predicting the deployment of the enemy forces will be beneficial to our combat. Based on the situation data of wargame, this paper proposes a method for predicting the unknown operator position of the enemy by training Graph Neural Networks. Firstly, after preprocessing the data, the transformation from situation to graph structure data is realized, and the graph structure data set of wargame situation is constructed for training Graph Neural Networks. Secondly, according to the characteristics of wargame situation and its data, GraphVAE model is modified to complete the graph structure data of wargame situation. Finally, a method for calculating the position of the enemy operator based on the completed graph structure data is designed. The effectiveness and feasibility of the proposed method are verified by experiments.

Keywords

• situation prediction|graph structure data|graph neural network|wargame situation