Deep Learning-Based Multipath Mitigation for Multi-Frequency GNSS Receiver in Challenging Environments

Abstract

In the last few decades, the space segment of the global navigation satellite systems (GNSS) has been expanding with new constellations and signals. The US Global Positioning System (GPS) has added signals on the L1, L2, and L5 carrier frequencies. The European Union (EU) Galileo GNSS is approaching a full operational status, and it has signals on three carrier frequencies (E1, E5, and E6). The Chinese BeiDou GNSS has achieved a global coverage, and it has signals on three carrier frequencies (B1, B2, and B3). The user segment has been slower in utilizing the available GNSS systems and signals due to the size and cost restrictions on wireless devices. However, many smartphones currently have at least dual-frequency reception. Beside the legacy L1 C/A signal, the L5 signal has emerged as a favorite for manufacturers.

GNSS positioning in multipath environments still suffers from the multipath problem, which remains the main threat to the application of the GNSS technology in civilian urban environments. This is because multipath signals are responsible for high positioning errors and they are complex to detect and mitigate. This talk will discuss the design of multipath mitigation algorithms for a multi-frequency GNSS receiver. The talk will explore the design challenges and investigate strategies to overcome these challenges.

The presented algorithms will focus on the L1 C/A and L5 signals of GPS. The L5 signal has enhanced structure and properties compared to the legacy L1 C/A signal. The length of the PRN code of the L5 signal is 10 times that of the L1 C/A signal and the chipping rate of the L5 signal is 10 times faster. This translates to a one chip of the L5 signal covering about 30 meters (m) as opposed to the nearly 300 m covered by one chip of the L1 C/A signal. However, the L5 signal is not immune to multipath errors and could give a degraded positioning accuracy when coupled with algorithms that depend on the reception of multipath signals to aid in the position estimation process, like some 3D mapping aided (3DMA)-based algorithms. Therefore, the aim is to utilize information that can be extracted from the tracking module of a GNSS receiver to build robust deep learning-based multipath mitigation (DNN-MP) algorithms. The DNN-MP algorithms are integrated into a 3DMA-based multi-frequency optimized position estimation (OPE-MF) algorithm. The results show a promising positioning performance.

Speaker

Prof. Nesreen I. Ziedan is a Professor and the Department Chair of the Computer and Systems Engineering Department, Faculty of Engineering, Zagazig University, Egypt. She received a PhD degree in Electrical and Computer Engineering from Purdue University, USA. She has several U.S. Patents, a book, and many papers on GNSS receiver design and processing. She is an IEEE Senior Member, and a member of the editorial board and a reviewer in electrical engineering and navigation-related journals.