A Neural Network-Based Intelligent Decision-Making in the Air-Offensive Campaign with Simulation

Abstract

Combat units in joint operations have huge decision space and many uncertain factors. Generally speaking, most of the traditional decision-making methods are based on rules, and it is impossible to establish a reliable mapping relationship between decision space and combat results. To promote the research of intelligent decision making in joint operations, the Equipment Development Department of the Central Military Commission held a challenge called ‘strategic plans on the computer, joint intelligent win’. In this challenge, the forces and the performance equipment are fixed at both the sides of the attack and defense. This setup helps the intelligent deci-sion-making agents to identify the scenarios which score high and have good learning scope in decision making. In the study, we propose an air offensive operations decision-making agent based on a neural network. To perform testing and analysis, we have used the neural network dataset available at a decision space. The decision space comprises of different decision-making rules and rando disturbances. The proposed model shows better results as compared to traditional rule-based operations and military expert decision-based operations in the test set.