Cluster Target Tracking Based on Multi-Sensor Adaptive GLMB Filter

In complex detection environments, unknown detection probability and clutter rate hinder accurate tracking of cluster targets. To address this issue, this paper proposes a novel multi-sensor adaptive generalized labeled multi-Bernoulli (MS-AGLMB) filter. Specifically, we consider interactions among cluster members and adopt a virtual leader–follower model to describe cluster kinematics. Given unknown environmental parameters, we employ an adaptive cardinalized probability hypothesis density (CPHD) filter to estimate the detection probability and clutter rate in real time. Furthermore, we use Gibbs sampling to efficiently truncate GLMB association hypotheses, obtaining the posterior density and solving the multi-sensor measurement partitioning problem. A joint prediction and update strategy enables simultaneous estimation of target trajectories, detection probability, clutter rate, and cluster structure. Simulation results demonstrate that the proposed algorithm achieves greater robustness in scenarios with time-varying detection probability and clutter rate, outperforming comparison filters in cluster target tracking.