A bayesian approach to multiple target localization

In this paper a multiple target localization problem is considered with only a partially known signal propagation model. Specifically, we assume that localization is to be effected by measuring the received signal strength (RSS) at each sensor. That RSS is modeled by a standard signal propagation model, though with unknown parameters. We adopt a Bayesian approach to propose a Markov Chain Monte Carlo (MCMC) type of algorithm for simultaneously estimating these unknown parameters and the source locations. Our approach also yields a posterior density function of these quantities conditioned on the RSS measurements. Such a density is useful for a visual inspection of the terrain to ascertain the source locations. The convergence of the algorithm is established under mild assumptions. Simulation results that support the analysis are provided.