UWB Localization Systems and Machine Learning

Research summary

Radio frequency (RF) signals are widely used for indoor localization, such as Bluetooth, Wi-Fi, and radio-frequency identification (RFID)-based localization systems. However, it is challenging for these RF technologies to achieve sub-meter level accuracy and robust localization. The ultra-wideband (UWB) signal can enable high-accuracy localization considering its following unique advantages, including sub-nanosecond-level time stamps, strong resistance to MPCs, and high throughput.

We are focusing on the following research problems:

NLOS error mitigation based on self-training learning in RF environment;
Ultra-wideband array localization system using wrapped PDoA;
Single-anchor localization system based on ultra-wide bandwidth signals;
Bayesian modeling for domain data and blind single-image super-resolution;
UWB multipath component identification and SLAM;
Machine learning algorithm on NLOS error mitigation.

Representative works

We propose a self-training learning scheme for RF environment, and design a multi-sensory information fusion localization system based on inertial sensors, map and UWB. The NLOS ranging error is reduced by more than 75% and results in 90th percentile of final localization error achieving 0.5m in a complex indoor environment.
- Y. Huang, S. Mazuelas, F. Ge, and Y. Shen, ’'Indoor localization system with NLOS mitigation based on self-training,’’ IEEE Trans. Mobile Comput., 2022, Early Access.

We develop a SAL system which achieves 3D high-accuracy localization using time and wrapped phase measurements of UWB signals, and achieve decimeter-level localization accuracy in outdoor and indoor environments.
- F. Ge and Y. Shen, ’'Single-anchor ultra-wideband localization system using wrapped PDoA,’’ IEEE Trans. Mobile Comput., 2022, Early Access.
- F. Ge and Y. Shen, “Real-time indoor localization on smartphones by multi-grained grid-based filters,” in Proc. IEEE Global Telecomm. Conf., Abu Dhabi, United Arab Emirates, Dec. 2018, pp. 1-6.

A single-anchor localization method that achieves high-accuracy multi-agent localization with high efficiency is proposed. The proposed method can achieve decimeter-level accuracy for multiple agents using three messages. Our method provides design guidelines for high-accuracy and high-efficiency multi-agent localization systems.
- T. Wang, H. Zhao, and Y. Shen, “An efficient single-anchor localization method using ultra-wide bandwidth systems,” Appl. Sci., vol. 10, no. 1, p. 57, Jan. 2020.

We establish a data-driven wireless localization and sensing framework with deep inference techniques for complicated inference problems, which can theoretically leverage both model-driven and data-driven knowledge of the wireless system and efficiently exploit both positional and situational awareness of agents from raw sensing measurements in harsh environments.
- Y. Li, S. Mazuelas, and Y. Shen, ’'Data-Driven Soft Range Information Estimation with Context Variable Inference,’’ IEEE Trans. Wirel. Commun., to be submitted.
- Y. Li, S. Mazuelas, and Y. Shen, ’'A Variational Learning Approach for Concurrent Distance and Environment Inference,’’ IEEE Trans. Wirel. Commun., submitted.
- Y. Li, S. Mazuelas, and Y. Shen, Variational Learning GNN for Direct Positional and Situational Map Inference, to be submitted.

We propose a passive anchor assisted localization (PAAL) scheme, where the active anchor obtains TOAAOA measurements to the agent while the passive anchors capture the signals from the active anchor and agent. The proposed method fully exploits the time-difference-of-arrival (TDOA) information from the measurements at the passive anchors to complement single-anchor joint TOAAOA localization.
- H. Lu, T. Wang, F. Ge, Y. Shen, “Robust UWB Array Localization through Passive Anchor Assistance,” China Commun., vol. 18, no. 4, pp. 1-13, Apr. 2021.

We propose a multipath assisted perception and localization algorithm based machine learning. The proposed algorithm utilizes machine learning to achieve robust multipath parameter estimation, and combines with an unsupervised Bayesian filter network to implement a multipath-assisted SLAM algorithm. The proposed algorithm exploits the potential of machine learning and environmental perception
- J. Liu, T. Wang, Y. Li, C. Li, Y. Wang, and Y. Shen, ’'A transformer-based signal denoising network for AoA estimation in NLoS environments,’’ IEEE Commun. Lett., 2022, Early Access.
- T. Wang, J. Liu, and Y. Shen, ’'A robust single-anchor localization method with multipath assistance in NLOS environments,’’ in Proc. IEEE Global Commun. Conf., Madrid, Spain, Dec. 2021, pp. 1-6.