A novel fusion method of conflicting evidences for clustering wireless sensor networks

Data fusion technology is an efficient way to decrease network energy consumption and recognition uncertainty of single sensor node for clustering wireless sensor networks. However, Dempster's combination rule may induce illogical results, when the information from different intra-cluster nodes highly conflict due to the background noise or flaws of the sensor itself. Through analyzing all evidences collected by cluster header, a novel aggregation algorithm based on support degree coefficient and conflict intensity is proposed. In the method, conflict intensity between every two bodies of evidence was analyzed, which divides conflict probability into useful and useless information respectively. In order to weaken the effects of abnormal evidences on fusion result, the combination sequence is made to be descending sort according to total conflict intensity of evidence. Additive strategy is adopted to obtain the support degree coefficient of single focal element of evidence set, based on which the useful information is assigned to different certainty propositions respectively. Numerical example showed that the proposed algorithm can provide more reasonable results with good convergence compared with other several modified combination rules.