Road networks are often monitored by only a limited number of sensors, leaving many segments unobserved. Researchers from the Technical University of Munich and Imperial College London developed RADGCN, an AI approach that learns traffic relationships from reliable observations and extends them across the network. Across three benchmark datasets and different sparse-sensor settings, RADGCN improved traffic-state estimates by 2.97%–11.82% compared with leading methods.
The team published their study in Communications in Transportation Research ( https://doi.org/10.26599/COMMTR.2026.9640046 ).
Accurate traffic information is essential for congestion monitoring, traffic control, route guidance and infrastructure planning. Yet installing and maintaining fixed detectors across an entire road network is costly, leaving many road segments without direct observations. A research team from the Technical University of Munich and Imperial College London has developed a machine-learning approach that estimates network-wide traffic conditions from sparse sensor data.
The study, titled "Relation-diffusion augmented network-wide traffic state estimation under sparse sensor deployment," has been accepted for publication in Communications in Transportation Research. It introduces the Relation-Augmented Diffusion Graph Convolution Network (RADGCN), a relation-aware framework for estimating traffic states where measurements are unavailable.
Many graph-based traffic models rely primarily on the physical links between roads or construct adaptive relationships from incomplete inputs. That can make learned relationships unreliable for road segments with no sensor data. RADGCN takes a different approach: it first learns functional traffic relationships from observed roads, where the data are reliable, and then propagates those relationships to the broader network.
Learning relationships before propagating traffic states
The model combines temporal traffic signals from observed roads with graph-aware attention to identify functional similarities between road segments. Learnable bilateral diffusion kernels then extend this information to unobserved parts of the network. The resulting relation matrix complements the physical road topology as an additional transition operator, allowing traffic states to be estimated through both connectivity and observation-supported functional dependencies.
RADGCN also uses K-order neighbourhood subgraph sampling and an auxiliary edge-prediction task during training. These components help the model preserve local network structure when it learns from subgraphs rather than from a complete road network.
More accurate estimates with sparse sensor coverage
The researchers tested RADGCN on the METR-LA, PEMS-BAY and PEMSD7M benchmark datasets. Across different sparse-sensor settings, the method consistently outperformed state-of-the-art comparison models, with average improvements of 2.97% to 11.82% across mean absolute error, root mean square error and mean absolute percentage error.
Additional analyses showed that the relation-diffusion mechanism, relation-augmented feature propagation and K-order sampling each contributed to the improvement. The approach also maintained its performance across datasets and different levels of sensor sparsity.
More complete traffic-state estimates could help agencies monitor large road networks where dense sensor deployment is impractical. The study was conducted by Tao Guo, Ningkang Yang and Constantinos Antoniou at the Technical University of Munich, and Kevin Yu and Panagiotis Angeloudis at Imperial College London. Source code is available at https://github.com/tao2000guo/RADGCN.
DOI Link:
https://doi.org/10.26599/COMMTR.2026.9640046
About Communications in Transportation Research
Communications in Transportation Research was launched in 2021, with academic support provided by Tsinghua University and China Intelligent Transportation Systems Association. The Editors-in-Chief are Professor Xiaobo Qu, a member of the Academia Europaea from Tsinghua University, and Professor Xiaopeng (Shaw) Li from University of Wisconsin–Madison. The journal mainly publishes high-quality, original research and review articles that are of significant importance to emerging transportation systems, aiming to serve as an international platform for showcasing and exchanging innovative achievements in transportation and related fields, fostering academic exchange and development between China and the global community.
It has been indexed in SCIE, SSCI, Ei Compendex, Scopus, CSTPCD, CSCD, OAJ, DOAJ, TRID and other databases. It was selected as Q1 Top Journal in the Engineering and Technology category of the Chinese Academy of Sciences (CAS) Journal Ranking List. In 2022, it was selected as a High-Starting-Point new journal project of the "China Science and Technology Journal Excellence Action Plan". In 2024, it was selected as the Support the Development Project of "High-Level International Scientific and Technological Journals". The same year, it was also chosen as an English Journal Tier Project of the "China Science and Technology Journal Excellence Action Plan PhaseⅡ". In 2024, it received the first impact factor (2023 IF) of 12.5, ranking Top1 (1/58, Q1) among all journals in "TRANSPORTATION" category. In 2026, its 2025 IF was announced as 12.7, maintaining the Top1 position (1/66, Q1) in the same category.
From Volume 6 (2026), Communications in Transportation Research will be published by Tsinghua University Press on the SciOpen platform with the official journal website at https://www.sciopen.com/journal/2097-5023 . We kindly request that all new manuscript submissions be made through the journal's submission system at https://mc03.manuscriptcentral.com/commtr
