Consensus-based Page's test in sensor networks

Page's test is a well-known statistical technique to approach quickest detection problems, namely the detection of an abrupt change in the statistical distribution of a certain monitored phenomenon. Running consensus is a recently proposed signal processing procedure aimed at reaching agreement among the nodes of a fully flat network, and its peculiar feature is the simultaneity of two stages: that of acquiring new measurements by the sensors, and that of data fusion involving inter-sensor communications.In this paper we study a quickest detector based on the running consensus scheme, and compare it to a bank of independent Page's tests. Exploiting insights from previous studies, we propose closed-form analytical approximations of the performances of these detection schemes and address a comparison in terms of relative efficiencies. The approximated performance figures are then checked by simulation to validate the analysis and to investigate non-asymptotic scenarios.