Dempster-Shafer Theory Based Robust Sequential Detection in Distributed Sensor Networks

Abstract

We propose a distributed sequential detector based on the Dempster-Shafer Theory of Evidence. First, we introduce a novel rule for the basic probability assignment. This rule is based on the distribution of the likelihood ratio and is shown to yield better results than existing ones while at the same time avoiding counter-intuitive and contradictory probability assignments. Second, we use the Dempster-Shafer combination rule to design a distributed sequential detection algorithm. Third, we show how to robustify the algorithm against outliers by leveraging neighborhood communication.