Dual‐mode multiple‐target tracking in wireless sensor networks based on sensor grouping and maximum likelihood estimation

Abstract

The problem of tracking multiple mobile targets, using a wireless sensor network, is investigated in this paper. We propose a new sensor grouping algorithm, based on the maximum sensor separation distances (G-MSSD), for estimating the location of multiple indistinguishable targets, either jointly or individually, depending on the distances between the generated groups. The joint tracking algorithm is formulated as a maximum likelihood (ML) estimator and solved through a modified version of the well-known Gauss-Newton (MGN) iterative method. We propose two candidate initial guesses for MGN based on G-MSSD in joint tracking mode, while for the individual mode, the information of each group is used to estimate the location of only the corresponding target. The Cramer-Rao lower bound (CRLB) for the variance of the proposed ML estimator is derived, and the potential conditions for reducing the CRLB are presented. Since tracking efficiency is affected by poor estimates, we present two criteria to evaluate the quality of estimates and detect the poor ones. An approach is also proposed for correcting the poor estimates, based on additional initial guesses. We demonstrate the effectiveness and accuracy of our proposed dual-mode algorithm via simulation results and compare our results with the Multi-Resolution search algorithm