Cubature belief propagation for self-localization of wireless networks

Abstract

Self-localization capability is a key enabler for a diverse set of important wireless applications. This paper presents a low-complexity belief propagation (BP) algorithm, named cubature belief propagation (CBP), for self-localization of wireless networks. The core of the proposed CBP algorithm is a third-degree spherical-radical cubature rule, which makes it possible to numerically compute the nonlinear multivariate moment integrals encountered in BP algorithm. CBP is fully distributed over an unstructured networks and does not require a fusion center, thus suits for an-hoc deployment. Moreover, since the adoption of conventional message passing mechanism, CBP can estimate node locations and represent location uncertainties simultaneously, and thus can be robust to outlier measurement errors. Simulation results show that, for a decentralized, cooperative, dynamic self-localization problem in wireless networks, CBP can outperform nonparametric/particle-based BP in lowing computation and communication overloads.