GPS-Based Vehicle Moving State Recognition Method and Its Applications on Dynamic In-Car Navigation Systems

In order to effectively determine whether a vehicle is turning or not, we proposed a method to map arbitrary consecutive GPS heading information to 2 dimensional feature space. Then we applied K-means clustering algorithm to divide the feature space into 2 classes: going straight and turning. After that, we used supervised learning algorithm to analyze these labeled data and build a model to recognize vehicle moving state. The experimental results showed that the model built in this way has good generalization. Based on the above research achievement, we designed and implemented a vehicle moving state recognition learning system for dynamic in-car navigation systems and applied this learning system to the map-matching field. The improved map-matching algorithm was tested on a complex urban road network and the result showed that the new algorithm can significantly improve the performance of the junction match.