Accurate vehicle position estimation using a Kalman filter and neural network-based approach

Abstract

Accurate detection of vehicle position plays an important role in many intelligent transportation systems, especially vehicle-to-vehicle applications. In this paper, we propose an Extended Kalman Filter (EKF) based method to detect Global Positioning System (GPS) errors for such vehicle-based applications. A machine learning methodology is presented for Kalman filter parameter tuning with application to GPS error correction in vehicle positioning. We also present a model free neural network that is trained on past vehicle GPS trajectories to predict the current vehicle position. Experimental results on real-world data show that the proposed system is effective for detecting and reducing GPS errors. The machine learning algorithm for EKF parameter tuning can be implemented through in-vehicle learning, and the proposed GPS error detection method can be implemented for in-vehicle applications.