Research On Intention Recognition Technology In Battlefield Based On Improved Markov Model

The battlefield situation is complex and changing rapidly.Whether we can identify the enemy’s battlefield intention timely,accurately and efficiently in the face of complex environmental factors is related to the decision-maker’s grasp of the overall situation of the battlefield and affects the pattern and trend of the whole war.In this paper,firstly,the definition,research significance and application prospect of battlefield intention recognition are analyzed and elaborated in detail.The task of intention recognition is divided into tactical level and macro strategic level.Different intention types are explained and formalized respectively,and the corresponding recognition algorithms are studied.Finally,the recognition experiments of different intention types are carried out.For maneuver intention of tactical level,considering the irrationality of classical HMM model under strong assumption,a weighted observation HMM model is designed,and three problems of probability calculation,parameter learning and decoding of the model are given.Finally,based on the air combat case,the model algorithm is analyzed and verified.The experiment shows that the model can identify the tactical level maneuver intention well.For the macro intention,considering the confusion of strategy pattern,the HMM-KNN combination model is designed,and KNN model is added to the HMM model for secondary identification and optimization.Taking the real-time strategy game as the test environment,the macro strategy intention recognition is verified.To sum up,this paper focuses on the characteristics and difficulties of intention recognition in the battlefield environment,analyzes the methods of extracting battlefield elements and constructing the framework of intention recognition,sums up the operational intention in the complex battlefield,and puts forward corresponding algorithms for different types of intention recognition.Experiments show the accuracy and feasibility of the recognition algorithm.