Intelligent Data Fusion Technology For Vehicle Cyber-physical System

With the development of big data and Vehicular Cyber-Physical Systems(VCPS),multi-sensor data fusion technology has been further applied.However,the sensors in the VCPS are susceptible to environmental factors,resulting in uncertainties and high conflicts.How to get a reliable fusion result and decision from these low-quality data is an urgent problem to be solved.In order to achieve effective fusion of high-conflict data,a new high-conflict data fusion algorithm based on Fuzzy inference mechanism and Belief entropy under the framework of Dempster-Shafer(DS)evidence theory is proposed in this paper,which mainly includes the following key points:1)The degree of conflict between the evidences in this paper is measured from two angles,namely the fuzzy nearness and the correlation coefficient between evidences.In order to accurately and comprehensively express the relationship among the fuzzy nearness between evidence,correlation coefficient and evidence conflict degree,a new Fuzzy-based Similarity Measurements(FSM)is proposed in this paper.2)This paper proposes that the improvement of the evidence source model not only considers the conflict information between the evidences,but also the information volume of the evidence itself.The belief entropy of evidence is calculated in this paper to obtain the amount of information contained in the evidence.By combining fuzzy theory and belief entropy,the weight of unreliable evidence in evidence sources is further reduced to improve the quality of evidence sources.3)A novel high-conflict data fusion algorithm based on the proposed FSM model and belief entropy is proposed in this paper.Firstly,the FSM method is used to measure the similarity between evidences.At the same time,the information amount of the evidence is calculated by the Belief entropy,and the weights of the evidence are obtained by considering the similarity and the information amount of the evidence.In this way,the average evidence is calculated,and then the DS combination rule is used to obtain the joint unified result by integrating the average evidence multiple times.The fusion algorithm proposed in this paper not only measures the degree of conflict between evidences from multiple angles,but also considers the information source of the evidence source itself.Combining these two factors eliminates the influence of noise evidence in the evidence source and improves the quality of the evidence source model.Therefore,the modified evidence source not only avoids the occurrence of counterintuitive results through the combination of DS combination rules,but also improves the accuracy of the fusion results.4)A VCPS-oriented traffic congestion state recognition model is designed in this paper.At the same time,the high-conflict data fusion algorithm proposed in this paper is used to simulate the fusion of evidences of traffic congestion state descriptions of vehicles in VCPS.The experimental results show that the proposed method can be effectively applied to the traffic congestion state recognition model.