Research Of Data Fusion Algorithm Based On Fuzzy Dempster-shafer Theory Of Evidence

Vehicular Cyber-Physical System(VCPS)can fuse the data collected by different vehicle-bone sensors from the driving environment.Based on the feedback control,the useful information extracted from the VCPS are feedback to vehicles to making the driving easier and the response faster,safer and smarter.The sensor data collected by VCPS are often heterogeneous,timeliness,massive and conflict.Therefore,how to fusing these data effectively is the key technology of the research of VCPS.In is paper,after the analysis of the basic theory and methods of the multi-sensor data fusion,the Dempster-Shafer(DS)evidence theory and fuzzy set theory are studied to solve the conflict data fusion.The main contents of this paper include the following aspects:1)Based on belief function and plausibility function of DS evidence theory,a new probability transformation(BPFT)is presented to transfer the basic probability assignment to probability vector.Compared to the classical pignistic-based transform(PPT)and plausibility function-based transform(PFT),BPFT can take full advantage of the given information of systems and can get more reliable and reasonable probability vector.2)A new dissimilarity measurement method is proposed based on the fuzzy nearness and the correlation coefficient.Firstly,the presented BPFT transformation method is used to transfer the basic probability assignments to probability vectors;Then,the fuzzy nearness is utilized to indicate the degree of the conflict between the probability vectors and the correlation coefficient is introduced to reflect the divergence degree of the hypothesis that two belief functions strongly support;Finally,the integration of the fuzzy nearness and the correlation coefficient by the fuzzy t-conorm rule is used as the new conflict measure.Simulation results show that,compared to the exiting method,the proposed conflict measure can characterize two aspects of the conflict between two evidence and the results are more effective.3)The conflict data fusing method is developed based on the fuzzy evidence theory.Firstly,the proposed conflict measure method is employed to calculate the similarity between sources of evidence and the fuzzy similarity matrix is constructed to get the degree of supporting and confidence,which are used to determine the weighting factors of evidence;Secondly,reassign the conflicts to evidence withweighting factors by Shafer's discounting rule;Finally,the rapid combination rule based on the matrix analysis is adopted to fuse the discounted sources of evidence.Simulation results show that the proposed evidence combination method can solve problems of conflict data fusion and it not only has a good robustness but also has a faster integration speed.4)After the analysis of the necessity of data fusion in VCPS,a VCPS-based real-time traffic detection system is designed in this paper.The proposed data fusion algorithm is applied to achieve the integration of traffic information collect form the VCPS-based real-time traffic detection system,which can help users to make a real time path planning and enhance traffic safety.