Technical Program

Paper Detail

Paper Title Quickest Detection of a Moving Target in a Sensor Network
Paper IdentifierTH4.R6.3
Authors Georgios Rovatsos, University of Illinois at Urbana-Champaign, United States; Shaofeng Zou, University at Buffalo, United States; Venugopal V. Veeravalli, University of Illinois at Urbana-Champaign, United States
Session Quickest Change Detection II
Location Sorbonne, Level 5
Session Time Thursday, 11 July, 16:40 - 18:00
Presentation Time Thursday, 11 July, 17:20 - 17:40
Manuscript  Click here to download the manuscript
Abstract To be considered for the 2019 IEEE Jack Keil Wolf ISIT Student Paper Award. The problem of quickest detection of a moving target in sensor networks is studied. At some unknown time, a target emerges in the sensor network, and one of the sensors in the network is affected, whose data generating distribution undergoes a change. It is assumed that as the target moves around in the sensor network, the sensor that is affected by the target changes with time. Specifically, if a sensor becomes unaffected, then its data generating distribution changes back to the pre-change mode. A discrete time Markov chain is used to model the location of the affected sensor, and thus the data generating distribution of the sensor network after the target emerges is a hidden Markov model. The goal is to detect the existence of the target as quickly as possible subject to false alarm constraints. A windowed test based on a generalized likelihood ratio approach is constructed, and its asymptotic optimality is further established. Numerical results are provided to demonstrate its performance.