Battlefield Agent Decision-Making Based on Markov Decision Process

Abstract

Battlefield decision-making is an important part of modern information warfare. It can analyze and integrate battlefield information, reduce operators’ work and assist them to make decisions quickly in complex battlefield environment. The paper presents a dynamic battlefield decision-making method based on Markov Decision Processes (MDP). By this method, operators can get decision support quickly in the case of incomplete information. In order to improve the credibility of decisions, dynamic adaptability and intelligence, softmax regression and random forest are introduced to improve the MDP model. Simulations show that the method is intuitive and practical, and has remarkable advantages in solving the dynamic decision problems under incomplete information.