A data fusion system of GNSS data and on-vehicle sensors data for improving car positioning precision in urban environments

Abstract

Accurate car positioning on the Earth's surface is a requirement for many state-of-the-art automotive applications, but current low-cost Global Navigation Satellite System (GNSS) receivers can suffer from poor precision and transient unavailability in urban areas. In this article, a real-time data fusion system of absolute and relative positioning data is proposed with the aim of increasing car positioning precision. To achieve this goal, a system based on the Extended Kalman Filter (EKF) was employed to fuse absolute positioning data coming from a low-cost GNSS receiver with data coming from four wheel speed sensors, a lateral acceleration sensor, and a steering wheel angle sensor. The bicycle kinematic model and the Ackerman steering geometry were employed to particularize the EKF. The proposed system was evaluated through experimental tests. The results showed precision improvements of up to 50% in terms of the Root Mean Square Error (RMSE), 50% in terms of the 95th-percentile of the distance error distribution, and 75% in terms of the maximum distance error, with respect to using a stand-alone, low-cost GNSS receiver. These results suggest that the proposed data fusion system for car vehicles can significantly reduce the positioning error with respect to the positioning error of a low-cost GNSS receiver. The best precision improvements of the system are expected to be achieved in urban areas, where tall buildings hinder the effectiveness of GNSS systems. The main contribution of this work is the proposal of a novel system that enables accurate car positioning during short GNSS signal outages. This advance could be integrated in larger expert and intelligent systems such as autonomous cars, helping to make self-driving easier and safer.