Robust and adaptive CKF-based adaptive parallel interacting multiple-model tracking algorithm for complex maneuvering targets

For the problem of tracking complex maneuvering targets, traditional interacting multiple-model algorithms suffer from reduced accuracy and slow convergence due to their reliance on prior information and the limited coverage of candidate models. This paper proposes an Adaptive Parallel Interacting Multiple-Model (APIMM) tracking algorithm based on a Robust Adaptive Cubature Kalman Filter (RACKF) driven by an innovation-norm criterion. First, an adaptive-parameter CS-Jerk (APCS-Jerk) model enhanced through hypothesis testing is developed to improve responsiveness to abrupt maneuvers. Second, an adaptive transition probability matrix is designed, along with a correction mechanism based on discrepancies between historical and current model probabilities, thereby improving the flexibility of model switching. Furthermore, an innovation-norm-based RACKF is formulated to achieve both adaptivity and robustness, and it is embedded within the APIMM framework. A multi-model set comprising the Singer and APCS-Jerk models is employed to achieve accurate matching across maneuvering regimes. The simulation results show that, in different noise conditions, the new algorithm cuts position error by at least 37.9% and velocity error by at least 24.0% against older methods.