A Low Complexity Gaussian Parametric Message Passing Based Cooperative Localization Algorithm

Based on the theory of factor graph and belief propagation, a low complexity cooperative localization algorithm with Gaussian parametric message passing is proposed to improve the performance where the non-cooperative localization methods failed due to the insufficient coverage of anchors. The system model is established according to the Bayesian rule. Weighted samples are used to represent the salient characteristics of the local message, and a Gaussian parametric message passing rule is designed to reduce the burden of the network traffic. By constructing a relative spatial relationship between the target and its neighbour nodes, a novel message initialization method is put forward to concentrate the samples where the messages have significant mass. In order to facilitate efficient computation of peer-to-peer messages, the nonlinear observation equation is linearized approximately by exploiting the Taylor expansion. Then the expression of the message updating is deduced and the detailed flow of the algorithm is shown. Simulation results show that the proposed algorithm leads to an excellent performance at the communication overhead and computational complexity, with losing negligible localization accuracy.