Localization with Noisy Android Raw GNSS Measurements

TL;DR

This paper investigates localization using noisy Android raw GNSS measurements, applying filtering techniques like MHE, EKF, and RTS to improve accuracy, with RTS achieving the best noise suppression and significant error reduction.

Contribution

It introduces the application of advanced filtering methods to enhance Android GNSS localization accuracy amidst hardware noise constraints.

Findings

RTS smoother reduces horizontal errors by up to 76.4% in static scenarios.

Filtering techniques significantly improve localization accuracy over baseline methods.

Experimental results validate the effectiveness of noise suppression methods.

Abstract

Android raw Global Navigation Satellite System (GNSS) measurements are expected to bring smartphones power to take on demanding localization tasks that are traditionally performed by specialized GNSS receivers. The hardware constraints, however, make Android raw GNSS measurements much noisier than geodetic-quality ones. This study elucidates the principles of localization using Android raw GNSS measurements and leverages Moving Horizon Estimation (MHE), Extended Kalman Filter (EKF), and Rauch-Tung-Striebel (RTS) smoother for noise suppression. Experimental results show that the RTS smoother achieves the best positioning performance, with horizontal positioning errors significantly reduced by 76.4% and 46.5% in static and dynamic scenarios compared with the baseline weighted least squares (WLS) method. Our codes are available at https://github.com/ailocar/androidGnss.