DISTRIBUTED MULTIPLE GAUSSIAN FILTERING FOR MULTIPLE TARGET LOCALIZATION IN WIRELESS SENSOR NETWORKS

Indoor target tracking appears in several engineering problems and is a key enabler to a myriad of new applications. Localization in such global navigation satellite system (GNSS)-denied environments typically relies on the use of existing infrastructures and already deployed technologies. In this paper, we are interested in received signal strength (RSS)-based multiple target tracking (MTT) in wireless sensor networks (WSN). From an estimation standpoint, two problems arise: i) standard Bayesian filtering techniques are not able to cope with high-dimensional systems, and ii) WSN are typically built with resource-constrained low-cost sensors, which implies the need for distributed algorithms. A possible solution is to use a multiple Bayesian filtering approach, where the state-space is partitioned in several lower dimensional subspaces, and then a set of parallel filters are used to characterize the marginal subspace posteriors. In this work, we propose a new distributed multiple Gaussian filtering (MGF) formulation, to solve both the curse-of-dimensionality in high dimensional systems and the need of distributed algorithms in network localization applications.