GNSS windowing navigation with adaptively constructed dynamic model
Zebo ZhouBofeng Li · 2015
The conventional dynamic model in the Kal- man filtering-based GNSS navigation usually contains the state information of only one previous epoch, which can hardly reflect the real complex dynamic characteristics of motion behaviors. To improve the adaptability and accu- racy of GNSS applications, a window-based polynomial fitting method is used to construct the dynamic model. In the given window with multiple state epochs, all candidate dynamic models with different model orders are self-con- structed by using these multiple state epochs in real time. Then, based on the model selection theory, a model eval- uation criterion is derived in the Bayesian framework to choose the optimal dynamic model. With this optimal constructed dynamic model, the improved navigation solution then can be obtained by a window-recursive approach. We test the proposed strategies by using the simulation and real GNSS vehicular experiments. All results demonstrate the validity and efficiency of the pre- sented method.