Network localization and navigation using cubature belief propagation

Network localization and navigation gives rise of a variety of new applications that relay on position information of nodes. In this paper, a belief propagation based Bayesian solution for network localization and navigation is presented. First, the state estimation problem of network localization and navigation is presented and formulated as a Bayesian inference problem, where the state of each node can be obtained by marginalizing the corresponding joint posterior probability density function (PDF). Then, taking advantage of the joint posterior PDF factorization, cubature belief propagation (CBP) is introduced to solve this marginalization, where Gaussian "messages" are iteratively generated and propagated through neighboring nodes, and a third-degree spherical-radical cubature rule is used to numerically compute the nonlinear multivariate moment integrals therein. Finally, network localization and navigation simulations using different BP algorithms shows that the proposed CBP based algorithm can achieves a nice tradeoff between estimation accuracy and communicational/computational complexity.