| Target threat assessment is a key issue in the collaborative multi-target attack. Itis a comprehensive assessment activity based on the qualitative and quantitativeanalyses of target, which can provide significant evidence for the commander tomake firepower allocation. In case of modern high-tech air defense combatenvironment, threat assessment of air target must be accurate and rapid. Aninaccurate judgment will lead to mistakes in decision-making and impact efficiencyof air defense operations. Adversely, prompt judgments affect the opportunity ofbattle. This dissertation provides an in-depth technical study of threat assessmenttechnology in the offensive and defensive of countermeasures of optoelectroniccountermeasure weapons using modern intelligent algorithms.The main contributions of this dissertation are described as follows:The basic theory of information fusion and threat assessment are introduced,and research and development status of threat assessment technology aresummarized. Workflow of threat assessment module and guidelines of threatsequencing are given according to application analyses of decision-making tasks andassistance decision-making in command and control system.The in-depth research of classic intelligent algorithms, such as differentialevolution (DE), biogeography-based optimization (BBO), cuckoo search (CS),particle swarm optimization (PSO), bat algorithm (BA), and firefly algorithm (FA), is conducted. On this basis, several new hybrid algorithms, like DE/CS (differentialevolution/cuckoo search), HS/BA (harmony search/bat algorithm), MFA (modifiedfirefly algorithm), and BAM (bat algorithm with mutation), are initially proposed incombination with other intelligent optimization technology. The experimental resultson benchmarks show that these novel methods significantly improve theperformance of the original methods.To improve the accuracy and usefulness of the target threat assessment in the aircombat, a target threat assessment model and algorithm based on Elman_AdaBooststrong predictor is originally proposed. Firstly, the Elman_AdaBoost strongpredictor is introduced; secondly, a target threat assessment model based onElman_AdaBoost strong predictor is established; at last, an algorithm is described.The experimental results show that the prediction error for Elman_AdaBoost strongpredictor algorithm is notably lower than the weak predictor.Moreover, to further enhance the accuracy and feasibility of the target threatassessment in the aerial combat, another target threat assessment model andalgorithm based on back-propagation (BP) neural network optimized by glowwormswarm optimization (GSOBP) algorithm is initially proposed. In GSOBP, GSO isused to simultaneously optimize the initial weights and thresholds of BP neuralnetwork. Target threat database is adopted to study the target threat predictionperformance of GSOBP, and the proposed method is also compared with BP andPSO_SVM. The experiment indicates that GSOBP has higher target threat predictionaccuracy than the normal BP and PSO_SVM.A variant of wavelet neural networks (WNN)-MWFWNN network is firstlyproposed to solve threat assessment more accurately, efficiently and effectively inthe aerial combat. How to select the appropriate wavelet function is adifficult-to-solve problem when constructing wavelet neural network. Thisdissertation proposes a wavelet mother function selection algorithm in terms ofminimum mean squared error, and then constructs MWFWNN network using theabove algorithm. Firstly, a wavelet function library is established; secondly, wavelet neural network is constructed with each wavelet mother function in the library, andupdate wavelet function parameters and the network weights according to therelevant modifying formula. The constructed wavelet neural network is trained withtraining set, and then optimal wavelet function with minimum mean squared error ischosen to build MWFWNN network. Experimental results show that the meansquared error is1.23×10-3, which is better than WNN, BP and PSO_SVM. |
PDF Full Text Request