Methods Of Threat Assessment For Aerial Targets Based On Cloud Model Theory

In order to achieve the operation superiority, the commander should estimate the enemy’sthreat quickly and exactly according to the situation of both sides in the operational commandprocess. However, because of the complexity of modern high-tech war and the uncertainty oftarget information obtained from object assessment, it is too difficult for a single commander togive an exact threat assessment quickly. Consequently, intelligent methods of threat assessmentare much in demand for assistant decision-making. Aiming at the uncertainty of aerial targetsagainst a background of joint air defense operations, the target threat assessment attributes areexpressed using cloud model, which combines fuzziness with randomness. The thesis presentstwo methods of threat assessment for aerial targets based on cloud model theory starting fromtwo different ideas, that is, logic reasoning and bayesian network. The contributions of this thesisare listed in three aspects as follows:(1) Threat assessment attributes set of aerial targets is built based on analysis of capability,intent and opportunity, the qualitative attributes are quantified, and the quantitative attributes arenormalized. Three layers of subject relation are defined, such as cloud group, cloud family, andcloud, and the threat assessment attribute cloud model of aerial targets is established as eachattribute is transformed into cloud model, by which the antecedent cloud generators are devised.(2) A novel method of threat assessment is put forward based on cloudy MIN-MAX centerof gravity reasoning by introducing cloud reasoning technique into the field of threat assessment.Firstly, the antecedent cloud generators are initialized according to the threat assessment attributecloud model of aerial targets. Secondly, the reasoning rules are made according to professionalknowledge and cloudy MIN-MAX center of gravity reasoning algorithm is designed, beforeattribute drops are obtained by inputting the normalized attribute value to correspondingantecedent cloud generators, and then a threat grade drop is got via step by step reasoning fromthe bottom up. Finally, reasoning repeatedly to eliminate the influence of target informationuncertainty on the overall threat grade, and the threat grade drops obtained from reasoning areinput to a backward cloud generator, regarding the output expectation as the final threat grade.The validity of the method is checked by simulation of threat assessment for twenty typical aerialtargets.(3) A method of threat assessment is proposed based on cloudy bayesian networkconstructed by combining cloud model with bayesian network. Firstly, the bayesian networkstructure is designed according to the background, and continuous observation node istransformed into cloud model, unifying the network into discrete bayesian network, i.e. cloudy bayesian network. Secondly, the condition probability table is ascertained according to objectiveknowledge and professional experience, or learning from sufficient effective samples. Thirdly,the observation variable value is input to the cloudy bayesian network, getting the probabilitiesof target belonging to each threat level. Finally, the final threat grade is obtained by probabilitycomposing and synthetical cloud creating after repeated inference. The validity of the method ischecked by simulation results.