Cooperative Localization and Target Tracking in Mobile Sensor Networks With Uncertainties in Measurement Models

This article studies the problem of cooperative localization and target tracking for mobile sensor networks in the presence of both Gaussian additive and multiplicative measurement noises. A group of more complex relative measurement models involving uncertainties, that is, the means and variances of the multiplicative noises being unknown, are considered. By introducing the Normal-inverse-Gamma distribution to model the unknown means and variances, a distributed algorithm is designed based on the variational Bayesian technique. Within the distributed framework, the proposed algorithm ensures that each robot can jointly infer its state and all targets' states, as well as the unknown noise statistics, by using only the local relative measurements and intermittent absolute measurements. Moreover, the covariance intersection fusion strategy is employed, which guarantees consensus among all robots on their inferred posterior distributions of each target's state. As a result, the proposed algorithm is independent on the true statistical characteristics of the multiplicative noises, which thereby gets rid of the limitation of the existing results requiring the means and variances to be precisely known. The effectiveness of the proposed algorithm is validated by simulation results.