Edge computing framework enables real-time tracking across large roadside sensor networks

Researchers have developed a real-time vehicle-tracking framework that can follow traffic across large roadside sensor networks with low latency and meter-level accuracy, a step that could help support safer and more responsive vehicle-road-cloud integration systems.

High-precision vehicle positioning and trajectory tracking are becoming increasingly important for intelligent transportation systems, especially as connected vehicles, roadside infrastructure, and cloud platforms are expected to work together in real time. But in large-scale road networks, the sheer volume of data, inconsistent sensor quality, and the need for very low latency make continuous tracking a difficult problem.

To address that challenge, the research team designed a distributed framework based on multi-access edge computing, or MEC, so that more processing can happen closer to the roadside sensors rather than relying entirely on centralized cloud computing. The framework combines multiple stages, including data preprocessing, calibration, multi-sensor trajectory matching, and trajectory prediction, and incorporates machine-learning methods such as LSTM-based prediction to improve performance in complex driving conditions.

According to the study, the system was able to continuously track tens of thousands of vehicles on highways while maintaining an average longitudinal error of 2.14 m, an average lateral error of 0.84 m, and an average speed error of 1.91 kph. The framework was evaluated on a large-scale road network with 1,777 sensors and continuous multi-vehicle tracking over 157 km of highway. Processing latency remained below 340 ms, suggesting that the approach can support near-real-time traffic perception at scale.

The authors also introduced evaluation metrics that consider factors such as trajectory smoothness and velocity consistency, aiming to provide a more comprehensive picture of tracking quality beyond simple positional error alone. Together, these design elements suggest that the framework is built not only for tracking accuracy, but also for stable and scalable deployment in complex roadside sensing environments.

If further validated in broader traffic scenarios, the framework could help improve road safety, traffic efficiency, and the responsiveness of connected transportation systems. Because the reported results come from experiments in large-scale highway settings, future work will likely be needed to assess performance across more heterogeneous urban environments and under a wider range of sensing conditions.

Reference

Author:

Yanbin Liu ^{a b}, Bolin Gao ^a, Peikun Lin ^b, Guangyu Tian ^a, Keqiang Li ^a

Title of original paper:

A framework for real-time vehicle tracking in large-scale roadside sensor networks

Article link:

https://www.sciencedirect.com/science/article/pii/S2773153725001124

Journal:

Green Energy and Intelligent Transportation

DOI:

10.1016/j.geits.2025.100362

Affiliations:

^a School of Vehicle and Mobility, Tsinghua University, Beijing 100084, China

^b Alibaba Cloud Computing, Hangzhou 310000, China

Image usage restrictions:

News organizations may use or redistribute this image, with proper attribution, as part of news coverage of this paper only.

Image credit:

GREEN ENERGY AND INTELLIGENT TRANSPORTATION