Reference acceleration-based dynamic model for maneuvering target tracking

A reference acceleration-based model combining the prior knowledge and dynamic maneuvering information was developed to accurately describe the dynamics of maneuvering targets. The acceleration is assumed an asymmetric distribution with a quadruple-uniform mixture probability density. The asymmetry adaptively describes the acceleration's variation based on the target's maneuvers. A discretized Kalman filter is used for the analysis. And the connections between the proposed model and some other representative models are expatiated. Simulations show that the model performs better than the Singer model and the "current" model when maneuvers occur, during typical target tracking process.