Enhancing GNSS Positioning Accuracy in Urban Canyons: A Multi-Epoch Residual-Based Consistency Check and IMU-GNSS Tight Coupling Approach

As urban populations continue to grow, the demand for precise positioning has become increasingly critical for various applications, including navigation, location-based services (LBS), transportation, delivery, and logistics. Global Navigation Satellite Systems (GNSS) are widely utilized for positioning; however, they encounter significant challenges in urban canyons, where signal obstructions and reflections can degrade positioning accuracy by tens to hundreds of meters.

To address these challenges, existing consistency check methods have been developed to detect and exclude erroneous observations from multiple GNSS constellations, thereby improving performance in open or low-density urban environments. Nevertheless, high-density urban areas present difficulties, as the majority of GNSS signals are often compromised by non-line-of-sight (NLOS) reception and multipath interference. While the integration of inertial sensors with GNSS technology has shown effectiveness in addressing GNSS outages, the accumulation of drift errors in pedestrian dead reckoning (PDR) still hinder performance in dense urban settings where GNSS solutions are consistently unreliable.

In this study, we propose a novel approach that tightly integrates PDR and GNSS data in the measurement domain to effectively identify fault-free measurements amidst a backdrop of contaminated signals. We introduce a multi-epoch smoothing algorithm designed to enhance positioning accuracy. Our method employs a two-stage consistency check algorithm to mitigate multipath effects, incorporating both satellite quality assessments and grid quality evaluations based on raw GNSS observations and inertial sensor data. Notably, we leverage time-series residuals from multi-epoch GNSS observations to identify fault-free measurements, moving beyond the limitations of single-epoch data. Additionally, grid quality is evaluated based on the discrepancies in residuals among high-quality satellites. To bolster the robustness and reliability of positioning, our algorithm integrates a positioning scheme that utilizes weight smoothing based on multi-epoch grid mapping and outlier mitigation through density-based spatial clustering of applications with noise (DBSCAN) clustering.

Field experiments conducted in typical urban environments in Hong Kong, utilizing a standard smartphone as the receiver, demonstrated substantial improvements over conventional consistency check methods and chip outputs. Our findings reveal that traditional consistency check methods underperformed compared to chip outputs in dense urban areas. In contrast, the proposed method significantly enhanced positioning accuracy across all trials, achieving accuracies ranging from 2m to 10m, compared to chip outputs that varied from 5m to 58m. The proposed approach yielded an improvement rate of 50% to 88% across different urban densities.

This innovative method is compatible with most consumer-grade devices, requiring no additional hardware, thereby offering enhanced convenience and intelligence for urban residents. Its ease of implementation across various brands and real-time operation with low computational load make it a versatile solution for improving positioning accuracy in complex urban environments.