Improving Vehicle Localization in Hard Environment Using GNSS-GSM Hybridization and Gaussian Mixture Noise

Abstract

In urban environment, Multipath and non-line-of-sight (NLOS) phenomena affect considerably the positioning accuracy. Today, GNSS-based positioning systems have penetrated several field, such as maritime transport and civil aviation requiring high positioning accuracy. For that, improving accuracy became a challenging work in several research laboratories around the world. Techniques such as hardware improvements and 3D city models can improve the GNSS performances in constrained environments, but usually show a high systems cost and can be impractical to implement. In order to ensure high accuracy positioning, we aim to explore the potential of Global System for Mobile Communications (GSM) signal powers combined with the GNSS (global navigation satellite system)signal. Our proposed approach aims to model the NLOS and Multipath errors by a Gaussian mixture distribution (GMM), in which GNSS and GSM are combined with PF to estimate vehicle position in a dynamic movement. To evaluate the proposed hybrid approach, a real GPS (Global Positioning System) signal is experimented through simulation. We obtain an improvement of 8 meter over the GNSS signal only without Gaussian Mixture modelization and GSM power and of 1 meter over the GNSS signal combined with GMM without GSM signal.