Group Target Tracking Algorithm Based on Integrated Group Structure Aided by the CBMeMBer Filter

‍ ‍Group target tracking has become increasingly important in various fields， such as military operations， autonomous vehicles， and low-altitude defense systems. A group target consists of multiple individual targets moving together at the same speed or in the same direction. Due to the advantages of random finite sets （RFS）-based filters in handling data associations with multiple targets， most current group target tracking algorithms are based on RFS filters. However， these filters often overlook important factors such as inter-target correlations and dependencies when tracking group targets. To address these factors， a novel group target tracking algorithm based on the cardinality balanced multi-target multi-Bernoulli （CBMeMBer） filter has been proposed. This algorithm integrates the group structure into the tracking process to improve the accuracy of target tracking. Initially， the algorithm estimates the structure of the target group using an adjacency matrix， treating the group of targets as an undirected graph. By evaluating the distances between individual targets， the algorithm determines the group target’s adjacency matrix and divides the group into multiple subgroups. Based on the motion states of targets within each subgroup， the algorithm classifies them into two categories： group centers and group followers. It then establishes motion equations accordingly. In the prediction step， the estimated group structure guides the prediction of target states. In the Gaussian mixture implementation step of the proposed algorithm， multiple Gaussian components are used to model the corresponding Bernoulli components. However， an excess of Gaussian components can lead to inaccurate group structure estimations. To address this issue， the algorithm prunes Gaussian components during the state extraction step. It retains only the Gaussian component with the highest weight for each updated Bernoulli component， enhancing the accuracy of the filtering process. Finally， simulation results demonstrate the algorithm’s effectiveness at achieving stable and precise group target tracking， and reveal its superior performance compared to traditional CBMeMBer filters.