Distributed expectation maximization in Appointment Rumor Routing in WSN

Abstract

One of the major problem in wireless sensor networks is query-driven routing, which especially arises when Sink searches for data for which the location in the network is unknown. Event information is propagated by Rumor Routing in some selected paths in the network, and as a result, an event trace is created. A route to the location of the event is established when a query agent crosses an event trace. The main objective of all Rumor-based algorithms is to increase the cross over rate for the query-agents. In Appointment-based Rumor Routing, contrary to other query-driven algorithms, agents' cross over occurs in a predetermined location called appointment point. In previous works, the appointment point is centrally calculated at the base station. In this paper, a distributed algorithm is proposed for appointment point selection. In order to estimate the parameters and the rank of a Gaussian Mixture used to model event locations, a Distributed Expectation-Maximization algorithm is designed to estimate the rank and parameters. In this case, a data fusion algorithm as well as a novel approach called Active Gaussian Mean is employed to determine the appointment point. The results of simulation under multiple scenarios provide a comparison of the proposed algorithms with the ones in the literature. The results show that the proposed algorithms are nearly optimal.