不完全建模下具有未知机动的任意形状集群目标自适应跟踪

Drone swarm, as an instantiation of embodied intelligence and a typical example of spatially indistinguishable group targets, is emerging as one of the main surveillance objects for air defense and anti-drone systems due to its low-cost, large-scale, and high-flexibility characteristics. Traditional methods for group target tracking rely on the fine-grained modelling of the group’s centroid motion and its contour. However, when the group target performs unpredictable maneuvers or exhibits an irregular geometric shape, these methods often suffer from unsatisfactory tracking errors due to the mismatch between the prior model and the actual one. Therefore, this paper proposes an adaptive tracking method for arbitrary-shaped group targets with unknown maneuvers based on incomplete modeling. First, the temporal maneuvering motion of the group is modeled using a nominal model combined with function fitting, while the Gaussian process is employed to depict the irregular group geometry. This establishes a spatio-temporal evolution model for the maneuvering group target, incorporating unknown weights. Second, basis points for the above weights are adaptively constructed using a low-discrepancy sequence to assist in describing the group’s maneuvering motion, which effectively decreases the computational cost of the subsequent filtering algorithm. Finally, an iterated extended Kalman filter, aided by the pseudo-measurements of centroid, is designed to obtain a precise and efficient recursive estimate of the maneuvering group’s centroid and contour, and the computational complexity of the proposed method is also analyzed. Simulation results for tracking “T”-shaped irregular group target with typical maneuvers demonstrate that the proposed method attains high-precision estimate results for the group centroid and shape while maintaining computational efficiency, in the case of unknown maneuvering.