Comparison and evaluation of advanced motion models for vehicle tracking

The estimation of a vehiclepsilas dynamic state is one of the most fundamental data fusion tasks for intelligent traffic applications. For that, motion models are applied in order to increase the accuracy and robustness of the estimation. This paper surveys numerous (especially curvilinear) models and compares their performance using a tracking tasks which includes the fusion of GPS and odometry data with an Unscented Kalman Filter. For evaluation purposes, a highly accurate reference trajectory has been recorded using an RTK-supported DGPS receiver. With this ground truth data, the performance of the models is evaluated in different scenarios and driving situations.