Study of Distributed Data Fusion Using Dempster's Rule and Cautious Operator

This paper presents new algorithms to process data exchanged in vehicular networks. In previous works, a distributed data fusion method using belief functions to model uncertainties has been proposed for smart cars network. Since the origin of data coming from other cars is unknown, this algorithm uses the idempotent cautious operator in order to prevent data incest. This operator has been proved to be efficient in the case of transient errors and ensures the fusion convergence. However, since the cautious operator is idempotent, the quantity of concordant sources does not change the result of fusion. Thus we propose several schemes adding Dempster's rule in order to improve the fusion when we can ensure that data come from independent sources. We introduce three new combinations layout of Dempster's rule and cautious operator and we compare them using real data coming from experiments involving several communicating cars in the context of the COMOSEF project.