Computationally Efficient Distributed Multi-Sensor Fusion With Multi-Bernoulli Filter

Abstract

This paper proposes a computationally efficient algorithm for distributed fusion in a sensor network in which multi-Bernoulli (MB) filters are locally running in every sensor node for multi-target tracking. The generalized Covariance Intersection (GCI) fusion rule is employed to fuse multiple MB random finite set densities. The fused density comprises a set of fusion hypotheses that grow exponentially with the number of Bernoulli components. Thus, GCI fusion with MB filters can become computationally intractable in practical applications that involve tracking of even a moderate number of objects. In order to accelerate the multi-sensor fusion procedure, we derive a theoretically sound approximation to the fused density. The number of fusion hypotheses in the resulting density is significantly smaller than the original fused density. It also has a parallelizable structure that allows multiple clusters of Bernoulli components to be fused independently. By carefully clustering Bernoulli components into isolated clusters using the GCI divergence as the distance metric, we propose an alternative to build exactly the approximated density without exhaustively computing all the fusion hypotheses. The combination of the proposed approximation technique and the fast clustering algorithm can enable a novel and fast GCI-MB fusion implementation. Our analysis shows that the proposed fusion method can dramatically reduce the computational and memory requirements with small bounded L1-error. The Gaussian mixture implementation of the proposed method is also presented. In various numerical experiments, including a challenging scenario with up to forty objects, the efficacy of the proposed fusion method is demonstrated. arXiv:1906.07991v1 [eess.SY]

On Arithmetic Average Fusion and Its Application for Distributed Multi-Bernoulli Multitarget Tracking

“…The proposed B2B-AA-Flooding, B2B-AA-Consensus, B2B-AA-CC and Standard AA-CC were compared with the noncooperative approach that does not carry out any internode communication and fusion. At first, we considered the GA-MB approach [49], [50] but it turned out that they (based on our implementation) are computationally intensive and do not work well with so many sensors as we consider here; indeed the GA fusion demonstrated degradation [17], [25], [34], [35] with the increase of the number of fusing sensors.…”

“…3) We propose the "target-wise fusion principle" to extend the AA fusion for fusing MBs, which divides the MB fusion problem into parallel Bernoulli-to-Bernoulli (B2B) fusion subproblems, each resolved by exact Bernoulli-AA fusion [42]. This result is significantly different from the existing MB-GA fusion [48]- [50] and also from existing AA fusion [17], [32], [34], [35], [37], [40]. 4) We investigate both the average consensus [21], [22] and flooding [51] algorithms for internode communication, proposing accordingly two multidimensional assignment (MDA) approaches for B2B association based on either pairwise set matching or multi-sensor data clustering.…”

Multi-Target Localization and Tracking Using TDOA and AOA Measurements Based on Gibbs-GLMB Filtering

“…Since the multi-target posterior probability density recursion requires the calculation of multi-set integral (24) and (25), its computational complexity is much larger than that of the single-target filtering process [16,61,62]. By the SMC method, the weighted particles can be estimated by recursive approximation to estimate the posterior probability density.…”

“…However, these filters can only obtain the scatter set estimation of the target but cannot form the target trajectories, though several heuristics have been proposed to join state estimates from different times steps to form trajectories. In spite of this, these filters have been widely used in many fields, for example, computer vision [20][21][22]; sensor scheduling [23,24]; multi-sensor fusion [25]. Reference [26] propose to solve the multi-target sensor management by using the random set method in the POMDP ] framework; References [27][28][29] use Cauchy-Schwarz divergence and Rényi divergence as information gains, respectively, provide a new sensor scheduling; robotics [30,31] and group target tracking [32].…”

A Novel Nonlinear Algorithm for Non-Gaussian Noises and Measurement Information Loss in Underwater Navigation

“…Because the stochastic non-Gaussian noise exists in signal, the noise information with low accurate may introduce the drifts of attitude and position. Furthermore, these methods cannot deal with the effect brought by system uncertainties [25]- [27]. Several algorithms based on EKF have been proposed, like adaptive EKF (AEKF) [28], fuzzy EKF (FEKF) [29], and the backing decoupling and adaptive EKF (BD-AEKF) [30].…”