Evidence Theory in WSN Application of Data Fusion
|School||Central South University|
|Course||Control Science and Engineering|
|Keywords||Data fusion D-S theory Clustering Aggregation timing control Aggregation tree rebuilding|
This paper researches on the application in the data fusion for wireless sensors network of D-S theory, one of the classic algorithm for conventional multi-sensors data fusion. The practical WSN usually needs to deal with uncertain information with data fusion technology for reasonable and effective results. D-S theory is a powerful and natural data fusion technology for the expression and fusion of uncertain information, thus it is significant for promoting the performance of WSN to research its application in the data fusion of WSN. Clustering and aggregation tree are the two main data dissemination strategies for the data fusion in WSN, thus this paper researches the application of D-S theory under these two strategies. In the data fusion of WSN, D-S theory can either play a leading role or supporting role. In the former, the data dissemination strategies are designed to achieve better fusion results of D-S theory, whereas D-S theory acts as a tool to evaluate data dissemination strategies and assist the protocol establishment in the latter.The first part of this paper introduces the model of multi-sensors data fusion and the concept of data dissemination in WSN and exemplifies some representative algorithm. And the next part introduces D-S theory’s basic concept and its simple operative model, moreover some solutions are also brought in proposed by former researchers for addressing the defects of D-S theory. Then the rest part of this paper proposes a WSN clustering algorithm for D-S theory, a fusion timing control algorithm based on the profit of data fusion and a topology rebuilding algorithm suitable to the WSN with data fusion. All algorithms proposed in this paper are simulated with OPNET and the simulation results show that these algorithms can achieve better performances comparing with former algorithms.