A Method Of Situation Assessment Based On Bayesian Network And D-S Evidence Theory

Information fusion with uncertainty is a hot issue in military field research.Information fusion technology can usually be divided into two levels, that is, lowlevel fusion and high level fusion, of which the former is position and attributeestimation of the target, while the latter conducts battlefield situation assessment andthreat estimation. The process of situation assessment starts from analyzing battlefieldsituation resulting from the acquired target and environment information of low levelfusion to finish the prediction of the change trend of battlefield. As a matter of fact,the process of situation assessment is to integrate multiple-source and multiple-typeinformation with uncertainty, and to finish the comprehensive inference of theseintegrated information, which contribute to the mastery of current and futurebattlefield situation. In this paper, we focus on situation knowledge representation andreasoning approaches of knowledge with uncertainty in complex battlefieldenvironment, and concrete research work is arranged as follows:Firstly, basic definitions and concepts of situation assessment are introducedbriefly based on analysis on information fusion knowledge and its functional model.Then related functional model of situation assessment and the presentation andinference of knowledge with uncertainty are described in detail, all of which provideus with an important theoretical foundation for the reasoning approaches mentioned inthis paper.Secondly, we introduce the construction and inference of Bayesian networkmodel, upon which we make it a clear the basic thoughts and application direction ofthe method of situation assessment based on Bayesian network. Furthermore, we putforward dynamic Bayesian network approach for situation assessment to adapt to thedynamic changes of battlefield situation which is a difficult problem for the traditionalBayesian network. The dynamic Bayesian network approach can provide real-timeestimate results with higher accuracy and better expression of the dynamic law ofbattlefield changes through converting static Bayesian network to dynamic Bayesiannetwork with time factor. The simulation results of examples of situation assessmentindicate the effectiveness and applicability of the dynamic Bayesian network.Thirdly, owing to the incomplete and imprecise information, and the diversity and uncertainty of the knowledge in military field, we give a research on battlefieldsituation assessment and inference based on D-S evidence theory. As an extendedmethod of Bayesian network, D-S evidence theory meet much weaker requirementsthan probability theory, which makes D-S evidence theory a better way in theexpression and fusion of uncertain information. The simulation results show that D-Sevidence theory can give superior reasoning results in battlefield situationassessment.Finally, HDSmP transformation method is introduced to address a specialproblem in proposition assignment, where the value of mass function is zero whenusing D-S evidence theory in situation assessment which couldn’t be well solvedthrough traditional probability transformation approaches. Through transforming thebelief function into probability function approximately, the HDSmP method canfinish situation assessment with higher accuracy in probability domain rather thanbelief domain. Examples are given to show the effectiveness of HDSmP method inreducing the uncertainty of situation assessment proposition.