This paper presents a noise adaptive variational Bayesian cubature information filter based on Wishart distribution. In the frame of recursive Bayesian estimation, the noise adaptive information filter propagating the information matrix and information state is derived. And the integration of recursive Bayesian estimation is approximated by cubature integration rule. Then, the inverse of measurement noise matrix is modeled as a Wishart distribution, so the joint distribution of posterior state and measurement noise can be approximated by the product of independent Gaussian and Wishart. Furthermore, the corresponding square root version is also derived to improve numerical characteristics. Simulation results with unknown and correlated measurement noise demonstrate the effectiveness of the proposed algorithms.

The traditional consensus-based filters are widely used in distributed sensor networks. However, they suffer from divergence when outliers occur. This paper proposes a robust consensus nonlinear information filter for distributed state estimation with measurement outliers. Unlike the Gaussian assumption in traditional consensus filers, the measurement of each sensor node is modeled here as a multivariate Student-t process with unknown parameters of the sufficient statistic. The variational Bayesian method is employed to jointly estimate the state and the parameters. As the state and parameters are coupled, the updated equation can be solved by fixed point iteration. The centralized outliers robust information filter is first derived for multiple sensors. It is then extended to a distributed version to fuse information from multiple interconnected local estimators. The integral of the consensus-based nonlinear filter is approximated by Gaussian approximation under the framework of the information filter. The consensuses are based on both likelihoods and prior probability distributions. The consensus and convergence of the proposed method are also analyzed. Simulation results show that the proposed approach is effective in dealing with outliers.

This article introduces a novel consensus-based labeled multi-Bernoulli (LMB) filter to tackle multitarget tracking (MTT) in a distributed sensor network (DSN), whose sensor nodes have limited and different fields of view (FoVs). Although consensus-based algorithms are effective for distributed fusion and MTT, it may be problematic when distributed sensor nodes have different FoVs. To deal with this issue, the proposed method constructs an extended label space mapping to overcome the “label space mismatching” phenomenon; after that, the model of the undetected multitargets is established so that the tracks can be initialized outside the FoV of local sensors; finally and most important, weight selection and evolution mechanism are proposed such that the fusion weights are automatically tuned for each track at each time step and consensus step. The efficiency and robustness of the proposed algorithm are demonstrated in a distributed MTT scenario via numerical simulations.

A labeled multi-Bernoulli (LMB) filter is presented to perform multi-target tracking (MTT) for the glint noise. The measurement noise is modeled as a multivariate Student-t process. The variational Bayesian method is applied in the LMB framework with the augmented state. The predictive likelihood is calculated via minimizing the Kullback-Leibler divergence by the variational lower bound. Simulation results show that our approach is effective in MTT with the glint noise

This paper introduces a novel consensus-based labeled multi-Bernoulli (LMB) filter to tackle multi-target tracking (MTT) in a communication resource-sensitive distributed sensor network (DSN). Although consensus-based approaches provide effective tools for distributed fusion and MTT, the requirement of iterative communication makes it impractical in resource limited situations. To deal with this issue, two event-triggered strategies are proposed and incorporated into the consensus-based LMB. Focusing on the information discrepancy between the local multi-target probability density function (PDF) and the time prediction of the latest broadcast one, the integral-triggering strategy (ITS) is introduced. Furthermore, by proving that the information discrepancy (Kullback-Leibler divergence) between two LMB densities with the same label space can be decomposed into the sum of the information discrepancy of each LMB component pair (LMB components with the same label), the separated-triggering strategy (STS) is proposed. The performance of the proposed algorithms is demonstrated in a distributed multi-target tracking scenario via numerical simulations.