Nonparametric belief propagation based positioning via distributed network formation

Nonparametric belief propagation (NBP) algorithm is a popular probabilistic localization method in wireless sensor networks. It is particle-based and can be applied in nonlinear and non-Gaussian inference problems. However, NBP has practical limitations in dense networks due to the high computational complexity and network traffic resulting from the ranging and information exchanges with neighboring nodes in cooperative localization. In this paper, we design a distributed network formation approach to select a sufficient number of beneficial links resulting in a new network for cooperative localization, which improves the efficiency of localization owing to the reduction of redundant links. In addition, we develop a metric to judge and filter the invalid NBP particles, which can increase the accuracy of the localization. Simulation results show that the proposed scheme outperforms conventional methods.