When a sensor can resolve the members in a cluster, it is difficult to accurately track each target due to cooperative interaction among the targets. In this paper, we research the tracking problem of resolvable cluster targets with cooperative interaction. Firstly, we use the stochastic differential equation to model the cluster coordination rules, and the state equation of the single target in the cluster is derived. On this basis, a Bayes recursive filter tracking method based on the combination of the DBSCAN clustering algorithm and the $\delta $ -GLMB filter is proposed. In the $\delta $ -GLMB filter prediction stage, the DBSCAN algorithm is used to determine the cluster where the target is located in real time. Then, the collaborative noise of the target is estimated, which will be used as the input to correct the prediction state of the target. The simulation and experiment results demonstrate the effectiveness of the proposed algorithm when the cluster is splitting, merging, and in reorganization.

High-resolution sensors can monitor and track multiple group targets and provide information support for battlefield situation assessment. However, the tracking accuracy of sensors is significantly impacted by the heavy-tailed noise with unknown characteristics. To solve the multi-group target tracking issue when the statistical characteristics of the noise are unknown, we propose a novel Poisson multi-Bernoulli mixture (PMBM) filter with nested Multivariate Myriad filter (MMF), which is based on the Gamma Student’s t inverse Wishart (GSTIW) mixture distribution of the state variables and is called the MMF-GSTIW-PMBM adaptive filter. First, we model the noise and group target extended states using the Student’s t-distribution and the inverse Wishart distribution, respectively, and build a group target tracking model based on finite set statistics (FISST) and PMBM filter. Second, the MMF is embedded into the PMBM filter to estimate the characteristics of the innovation distribution and adaptively adjust the noise’s freedom and scale matrix parameters. Finally, to further enhance the adaptive estimation capability of the MMF, the multi-window weighted fusion technique is employed to optimize the selection of sampling windows. Simulation experiments show that the proposed filter is capable of adaptively estimating the noise characteristic parameters and accurately tracking multiple group targets. It has higher estimation accuracy and is more robust to noise than the GGIW-PMBM filter and GSTIW-PMBM filter.

The trajectory prediction of the hypersonic glide vehicle (HGV) can provide hit point information for early warning systems, which is of great significance for near‐space defence operations. However, the heavy‐tailed noise caused by abnormal environmental disturbance and abrupt changes in vehicle trajectory seriously affects the accuracy of HGV trajectory prediction. To solve the problem of trajectory prediction for HGV under heavy‐tailed noise, we propose an adaptive multivariate Student's t‐process regression method called aEM‐MVTPR. Firstly, a heavy‐tailed noise model is developed using the Student's t‐distribution. Secondly, a multivariate Student's t‐process regression (MVTPR) for HGV trajectory prediction is derived, and the method can learn the HGV trajectory time series features and mine the correlations among the trajectory variables. Finally, to further improve the method's robustness, we use the accelerated expectation maximisation algorithm and Pearson correlation analysis to adaptively estimate and adjust the initial values of the MVTPR parameters. The simulation experimental results show that the proposed method has more accurate predicted values of trajectory variables than multivariate Gaussian process regression (MVGPR) under heavy‐tailed noise conditions. In addition, the ability of the aEM‐MVTPR to adaptively adjust the parameter makes it more robust than the MVTPR under different noise environments.

Due to the limitation of sensor resolution, multiple target locations in close proximity can result in merged measurements, which makes it difficult for traditional measurement association assumptions to be met and makes the filter unable to resolve single target trajectories. In this paper, we use random finite set (RFS) theory to develop a novel trajectory Poisson multi-Bernoulli mixture (TPMBM) filter for partly resolvable multi-target (PRM) tracking, which is called the TPMBM-PRMT filter. First, we improve the likelihood model for the unresolved measurements and use the breadth first search (BFS) algorithm to partition the predicted set of detected trajectories. In addition, a trajectory subset-measurement association update method is proposed, where all trajectories within the subset share the same measurement, and a trajectory Poisson multi-Bernoulli (TPMB) approximation is designed to generate multiple high-quality association hypotheses and ensure the feasibility of the filter. Finally, the Gaussian L-scan approximation is used in the filter implementation, which reduces the computational cost while ensuring the filtering efficiency. Simulation results show that the filter in this paper can effectively handle the merging of multi-target measurements and achieve stable tracking of PRM.

This article proposes a filter based on the Poisson multi-Bernoulli mixture (PMBM) distribution for tracking point and extended targets with type conversion (PETC). First, we design a generalized measurement model based on the fixed measurement density. The model attributes target type conversion to changes in the projected extension of the target. Second, we systematically derive the PMBM recursion process in a single-target hybrid space, which allows the Gaussian density and the Gamma Gaussian inverse Wishart (GGIW) density to propagate simultaneously. We design a density transition model to capture possible conversions between these two densities. In addition, we create an expectation-maximization (EM) method to improve the estimation accuracy of the density transition probability. Lastly, a more efficient PMB approximation to the PMBM is derived. Simulation experiments demonstrate the superiority of our filters in maintaining continuous tracking when the target type is converted.