Publication

Abstract : As the automotive world is moving towards its ultimate aim of fully autonomous vehicles, predicting the trajectories itself and that of the neighboring vehicles is essential for each vehicle. This paper proposes a novel filtering method for predicting the path of an ego vehicle (i.e., a vehicle of interest) using measurements by a Global Positioning System (GPS) device and combining such measurements with multiple candidate kinematic motion models. The proposed Multiple Model filtering method adapts to the noise conditions as inferred by measurements. It is shown that the proposed adaptive method provides a considerable level of improvement compared to the existing non-adaptive multiple model filtering.