Interacting Multiple Model Particle-type Filtering Approaches to Ground Target Tracking

Ground maneuvering target tracking is a class of nonlinear and/or no-Gaussian filtering problem. A new interacting multiple model unscented particle filter (IMMUPF) is presented to deal with the problem. A bank of unscented particle filters is used in the interacting multiple model (IMM) framework for updating the state of moving target. To validate the algorithm, two groups of multiple model filters: IMM-type filters and particle-type multiple model filters, are compared for their capability in dealing with ground maneuvering target tracking problem. Simulation shows that particle-type filters outperform IMM-type filters in the estimate accuracy and the IMMUPF method relatively has much better performance than the IMMPF method.