Session A1: Alternatives, Backups, Complements to GNSS

Accuracy Analysis and Simulation of Angle of Arrival Estimation in 5G Positioning
Zihao Li, Yanhong Kou, Chao Sun, Honglei Qin, Tian Jin, Beihang University
Location: Beacon A

Global Navigation Satellite System (GNSS) is prone to degraded visibility or even complete failure in indoor or urban canyons, while 5G, as a system that currently provides both indoor and outdoor coverage, offers a possible supplementary means for User Equipment (UE) positioning under these conditions. The 3GPP TR 38.855 report points out a synchronization error exceeding 50ns may exist among different 5G Base Stations (BS), which leads to a distance error exceeding 15m in the commonly TOA/TDOA measurement for 5G positioning, ultimately resulting in an even larger positioning error. However, the introduction of technologies such as massive Multiple Input Multiple Output (MIMO) antennas, Ultra Dense Network (UDN) and millimeter wave (mmWave) has significantly improved the spatial angle resolution of 5G system. Based on 5G NR, accurate estimation of downlink Angle of Departure (AOD) and uplink Angle of Arrival (AOA) can be achieved at the BS side. The angle measurements combined with other measurements can provide enhanced positioning performance. However, 3GPP standards do not limit the specific angle estimation method employed by users. While its Cramer-Rao Lower Bound (CRLB) can be theoretically derived, the accuracy of angle estimation in real-world application environments remains unproven. At the same time, several super-resolution angle estimation and maximum likelihood estimation algorithms, utilizing array antennas, can be applied to 5G uplink AOA estimation, but related systematic analysis and comparison primarily focus on linear arrays, with limited studies exploring the planar arrays used in practice. Furthermore, there is a lack of joint analysis and comparison of the accuracy of angle estimation and positioning for specific positioning scenarios.
This study first simulates the antenna distribution on the 5G BS, and derives the CRLB for angle estimation as well as the CRLB based on AOA positioning. Specifically, the BS is deployed as a planar array with its antenna size configured according to the practical 5G BS. In addition, this study focuses on the positioning of static users or pedestrian handheld terminals, incorporating the recommendations of the 5G standard to configure the SRS-Pos signal sent by the UE. The uplink AOA measurement depends on this signal. To achieve the highest pilot density in the frequency domain, the minimum subcarrier setting in the comb pattern is periodically mapped to the physical resource in each slot, enabling the SRS-Pos signal to be transmitted over the entire signal bandwidth. The propagation channel primarily considers Line-of-Sight (LoS) and Non-Line-of-Sight (NLoS) channel, and the urban environment signal propagation path loss models in the 3GPP TR 38.901 protocol provide the signal power loss. In addition, the propagation delay is set according to the distance between BS and UE in LoS channel while the delay caused by NLoS is modeled as a commonly used exponential distribution.
After sequentially confirming the basic configuration, signal configuration, and channel configuration at the BS side, we conduct simulations of angle estimation under LoS/NLoS. With varying signal-to-noise ratio (SNR) conditions, we simulate multiple angle measurements, utilizing the beam search method provided by 3GPP, together with several super-resolution estimation algorithms based on multi-antenna, including the MUltiple SIgnal Classification (MUSIC), Estimation of Signal Parameters using Rotational Invariance Techniques (ESPRIT), and improved beam space MUSIC. In addition, referring to the typical station layout strategy in urban environments, we simulate specific positioning scenarios and evaluate the positioning performance for each angle estimation algorithm. Preliminary results and analysis show that the angle estimation errors of all the aforementioned methods exceed the CRLB, with angle measurement from the beam search method exhibiting lower resolution. MUSIC and other super-resolution algorithms can improve the accuracy of angle estimation, achieving a difference of no more than 3.5° compared to the CRLB, provided that the SNR is not less than 10dB. Notably the MUSIC algorithm based on the improve beam space demonstrates the smallest root mean square error (RMSE) in angle measurement, which is only 1.8° over the CRLB. In positioning simulations, when compared to beam search angle positioning, the super-resolution angle positioning can achieve a confidence level of 77% with its 3D positioning error within 10m assuming a BS distance of 100m, approaching more closely to the CRLB of AOA positioning.
In summary, we derive the CRLB for angle estimation based on planar arrays and analyze the accuracy of AOA estimation of different algorithms based on planar arrays in practical application environments. Furthermore, we derive the CRLB for AOA positioning and conduct Monte Carlo simulations to evaluate the positioning errors of various angle measurement algorithms in specific positioning scenarios in urban environments.