From autonomous vehicles to smart delivery systems, modern urban innovations rely on precise and timely location data. Fifth-generation (5G) mobile networks offer high speed and responsiveness, but urban structures—like buildings and overpasses—often block or reflect signals, creating non-line-of-sight (NLOS) errors. These distortions can push location errors into tens or even hundreds of meters. Though current solutions include filtering techniques and deep learning models, they often depend on large datasets or extra hardware, limiting their effectiveness in real-time deployments. Due to these persistent problems, there is an urgent need for positioning methods that can deliver accurate results using only 5G base-station data—especially in urban NLOS environments.
To solve this urban challenge, a team of engineers from Beijing Institute of Technology and China Mobile developed a new algorithmic framework, recently published (DOI: 10.1186/s43020-025-00165-w) in Satellite Navigation (April 2025). The method leverages virtual base-stations—software-generated nodes derived from real 5G measurements—to refine user location estimates without relying on external sensors. It integrates particle filters and the Augmented Dickey-Fuller (ADF) test to enhance accuracy while controlling computational costs. Field tests and benchmark datasets proved its advantage, showing it as a powerful and scalable alternative for high-precision 5G positioning.
The core of the algorithm lies in a three-part design: a Time-of-Arrival (TOA) positioning model, a virtual base-station generation method, and a stability-checking mechanism. First, the TOA model collects raw signal timing data from multiple base stations to estimate user position. Then, the VBS module refines this estimate by generating virtual base-stations through a particle filter coupled with a constrained random-distribution algorithm. This process adapts dynamically to local signal noise using known error propagation laws.
What makes the method distinct is its ability to recognize when these iterations have stabilized, using the Augmented Dickey-Fuller (ADF) test—commonly used in time-series analysis—to determine when further updates would be redundant. In both open-source datasets and a live urban testbed in Beijing, the algorithm demonstrated up to 21.09% improvement in positioning accuracy compared to traditional models, and outperformed the Levenberg–Marquardt algorithm by cutting computation time by over 75%. The residual errors were more concentrated and stable, especially in horizontal directions, which are critical for real-world navigation. This indicates that the method is not only accurate, but also highly efficient and applicable in mixed LOS/NLOS environments.
“Our goal was to tackle the signal chaos of the city with a lightweight, deployable solution,” said lead author Song Bao. “Virtual base-stations give us the flexibility to work with what we already have—raw 5G data—while smart filtering keeps our model efficient and stable.” Co-author Wang Bo added, “We believe this work opens the door to next-level location services in complex urban spaces, all without requiring new infrastructure or massive training datasets. It's a forward-compatible solution for future smart cities.”
This virtual base-station approach is poised to transform how urban environments handle 5G positioning. Because it enhances signal accuracy without needing additional sensors or large data training, it fits seamlessly into existing networks. This makes it ideal for applications such as autonomous vehicles, logistics robots, and emergency responders navigating complex cityscapes. Its real-time capabilities also support low-latency services like augmented reality and intelligent transportation. Future development will focus on refining the algorithm with more observational data types while preserving its fast, lightweight architecture—bringing high-precision 5G location services even closer to everyday reality.
###
References
DOI
10.1186/s43020-025-00165-w
Original Source URL
https://doi.org/10.1186/s43020-025-00165-w
Funding information
This study was supported by China Mobile Group Device Co., Ltd Fund (CMDC-202401967, CMDC-202402083).
About Satellite Navigation
Satellite Navigation (E-ISSN: 2662-1363; ISSN: 2662-9291) is the official journal of Aerospace Information Research Institute, Chinese Academy of Sciences. The journal aims to report innovative ideas, new results or progress on the theoretical techniques and applications of satellite navigation. The journal welcomes original articles, reviews and commentaries.