The paper addresses the problem of distributed multi-target tracking (MTT) in a network of sensors having different fields of view (FoVs). Probability hypothesis density (PHD) filters are running locally in every sensor for MTT. The weighted arithmetic average (WAA) fusion rule is employed to fuse multiple local PHD densities due to its computational efficiency. First, we provide a theoretical analysis showing that the standard WAA fusion among sensors with different FoVs is unsuitable from the perspective of the principle of minimum discrimination information (PMDI). In fact, the information inconsistency among sensors due to the different FoVs unavoidably leads to an underestimation of the fused PHD. Then, motivated by the analysis, we devise two novel approaches to address the different FoV issue. The first approach accounts for the case where the sensor FoVs are known and time-invariant. The second one deals with the more complicated case where the actual sensor FoVs can be cropped due to unknown line of sight obstructions or sensors can obtain the information outside their current FoVs because of platform movement or information feedback. The essence of the proposed approaches is to perform the WAA in a more robust way by employing a set of state-dependent fusion weights which are computed online. The Gaussian mixture implementations of the proposed methods are also presented. Various simulation experiments, including a challenging tracking scenario involving six sensors with different FoVs and also random line of sight obstructions, are designed to demonstrate the efficacy of the proposed fusion approaches.

Multiple extended target tracking (METT) is a common and challenging problem. Various solutions for METT have been proposed, however, most of them focus on the single-sensor or centralized multi-sensor scenarios. In this letter, we explore the multi-sensor METT problem in a distributed fusion framework. Specifically, there are two stages in the implementation process: 1) to perform a Gamma Gaussian Inverse Wishart Cardinalized Probability Hypothesis Density (GGIW-CPHD) filter for each sensor node, and 2) to perform a fusion by resorting to the so-called Generalized Covariance Intersection (GCI) fusion rule. In the fusion stage, we derive an approximate GGIW mixture form of the fused spatial density. Lastly, simulation experiments via a consensus sensor network are provided to verify the effectiveness of the proposed approach.

This paper proposes a robust Poisson multi-Bernoulli mixture (R-PMBM) filter immune to the unknown detection probability. In a majority of multi-object scenarios, the prior knowledge of detection probability is usually uncertain, which is often estimated offline from the training data. In such cases, online filtering is always unfeasible or unrealistic, otherwise, significant parameter mismatches will result in biased estimates (e.g., state and cardinality of objects). As a consequence, the ability of adaptively estimating the detection probability for a sensor is essential in practice. Based on the analysis, we detail how the detection probability can be estimated accompanied with the state estimates. Besides, the closed-form solutions to the proposed method are derived by means of approximating the intensity of Poisson random finite set (RFS) to a Beta-Gaussian mixture (BGM) form and density of Bernoulli RFS to a single Beta-Gaussian form, named BGM-PMBM filter. Simulation results demonstrate the effectiveness and robustness of the proposed method.

In this paper, we address the problem of resolvable group target tracking (RGTT) and the application of RGTT in the sensor control scenarios. The aim of our approach is to further improve the tracking performance of multi-Bernoulli (MB) filter by exploiting the group structure information. Firstly, a single-target state transition function (SSTF) is derived from the direct integral solution of the single-target state stochastic differential equation (SDE), which is used to revise the prediction process of the standard MB filter. Then, combing the Gaussian mixture (GM) implementation, the prediction process of the proposed SSTF-based MB (SSTF-MB) filter is detailed. Next, considering the benefits and importance of group structure in the sensor control scenarios, the proposed SSTF-MB filter is applied into a mobile sensor platform whose control actions (e.g., motion direction) are uncertain. Specifically, the proposed SSTF-MB filter is merged into a sensor control strategy where the Cauchy-Schwarz (CS) divergence is selected as the objective function to compute the discrepancy between two first-order moment approximations of MBs. Finally, the efficiency and performance of the proposed SSTF-MB filter and its application to sensor control scenario is well demonstrated in the simulation experiments.

The Poisson multi-Bernoulli mixture (PMBM) filter is conjugate prior composed of the union of a Poisson point process (PPP) and amulti-Bernoulli mixture (MBM). Considering that the single model is not enough to guarantee stable tracking performance for maneuvering targets, in this paper, a multiple model PMBM (MMPMBM) filter is proposed to cope with this problem. The proposed MM-PMBM filter extends the single-model PMBM filter recursion to multiple motion models by exploiting the jump Markov system (JMS). The performance of the proposed algorithm is examined from two scenarios with different detection probabilities. Moreover, the robustness of Markovian model transition probability matrices (TPMs) for the proposed MM-PMBM filter is also explored. The simulation results demonstrate that the proposed MM-PMBM filter performs well in terms of the tracking accuracy, including the target states and cardinality estimates, and also has good tolerance with respect to different TPMs.

This paper proposes a solution of distributed multiple resolvable group targets tracking (DMRGTT) by exploring the labeled multi-Bernoulli (LMB) filter and distributed fusion rules including generalized covariance intersection (GCI) and weighted arithmetic average (WAA). Due to the proximity of targets’ positions within the same group and the label mismatching problem for the fusion between LMB densities, only utilizing the similarity or distance between targets from different sensors cannot return a satisfactory solution of label matching (LM) results. To cope with this problem, the hypergraph matching (HM) theory is adopted to assist in finding the matched labels. Specifically, the hyperedge-to-hyperedge (H2H) matching matrix is first constructed and a vertex-to-vertex (V2V) matrix is returned. Then, the V2V matrix is further combined with the traditional distance matrix to construct a joint cost matrix. Next, substitute the joint cost matrix into the Hungarian algorithm and the LM results are obtained. Lastly, pairwise GCI-LMB and WAA-LMB fusion processes are performed based on the obtained LM results. Experiment results reveal the effectiveness and robustness of the proposed fusion approaches.