International audience

• This study interpolates urban street-level traffic volume • A graph-based model captures spatio-temporal traffic patterns • Method includes node features and custom loss function • Results are validated on novel Berlin and New York City datasets • Model outperforms baselines, especially under data scarcity Graph Neural Networks have shown strong performance in traffic volume forecasting, particularly on highways and major arterial networks. Applying these models to urban street networks, however, presents unique challenges: urban networks are structurally more diverse, traffic volumes are highly overdispersed with many zeros, spatial dependencies are complex, and sensor coverage is often very sparse. To address these challenges, we introduce the Graph Neural Network for Urban Interpolation (GNNUI), a model designed specifically for citywide traffic volume interpolation. GNNUI employs a designated masking strategy to learn interpolation, integrates node features to capture the different functional roles across the street network, and uses a loss function tailored to zero-inflated traffic distributions. We evaluate GNNUI on two newly constructed, large-scale urban traffic volume benchmarks, covering different transportation modes: Strava cycling data from Berlin and New York City taxi data. Across multiple evaluation metrics, GNNUI outperforms both the state-of-the-art graph-based interpolation model IGNNK and the widely used machine-learning baseline XGBoost, reducing MAE by at least 13% on Strava data and 7% on Taxi data while better capturing the empirical traffic distribution, and improving the identification of zero-traffic streets. Additionally, the model remains robust under the realistic case of extremely scarce ground truth sensor data. When sensor coverage is reduced from 90% to 1%, the MAE increases by approximately 48% on Strava and 76% on the taxi data, despite the near-complete removal of sensor information. We also examine how graph connectivity choices influence model performance, and find that a simple and computationally efficient binary adjacency matrix outperforms distance or similarity based ones.

Modeling in transportation is essential for planning and operating current and future mobility and logistics systems, with significant implications for modern economies. However, many of these modeling tasks involve computationally complex problems that challenge traditional models and forecasting techniques. Quantum computing offers a promising way to address these difficulties. This systematic literature review examines the state-of-the-art in applying quantum computing to transport modeling, including and classifying 40 studies, selected through a systematic search of three databases using predefined inclusion and exclusion criteria. We identify three main types of quantum computing techniques used: quantum algorithmic primitives, quantum optimisation methods, and quantum machine learning approaches. With respect to the transport modeling problems addressed, the reviewed works cover forecasting demand, systems’ operation economic variables, predicting network states, estimating vehicle dynamics, and queueing systems. A notable trend is the application of quantum walks and quantum neural networks for the prediction of network state variables. Our review also underscores key challenges and research gaps, including the lack of standardized benchmarks and theoretical guarantees, particularly for quantum machine learning, the underexploited use of quantum walks for vehicle dynamics estimation, and quantum annealing for network state forecasting. We conclude with future research directions aimed at closing these gaps and building stronger theoretical foundations for sustained progress.

Cycling is critical for cities to transition to more sustainable transport modes. Yet, safety concerns remain a critical deterrent for individuals to cycle. If individuals perceive an environment as unsafe for cycling, it is likely that they will prefer other means of transportation. Yet, capturing and understanding how individuals perceive cycling risk is complex and often slow, with researchers defaulting to traditional surveys and in-loco interviews. In this study, we tackle this problem. We base our approach on using pairwise comparisons of real-world images, repeatedly presenting respondents with pairs of road environments and asking them to select the one they perceive as safer for cycling, if any. Using the collected data, we train a siamese-convolutional neural network using a multi-loss framework that learns from individuals’ responses, learns preferences directly from images, and includes ties (often discarded in the literature). Effectively, this model learns to predict human-style perceptions, evaluating which cycling environments are perceived as safer. Our model achieves good results, showcasing this approach has a real-life impact, such as improving interventions’ effectiveness. Furthermore, it facilitates the continuous assessment of changing cycling environments, permitting short-term evaluations of measures to enhance perceived cycling safety. Finally, our method can be efficiently deployed in different locations with a growing number of openly available street-view images.

Through the vast adoption and application of emerging technologies, the intelligence and autonomy of smart mobility can be substantially elevated to address more diversified demands and supplies. Along with this trend, a systematic collaboration among three essential elements of smart mobility services, namely devices, data and functions, is being studied to comprehensively break down the intrinsic barriers that existed in current solutions, to support the integration of connectable devices, the fusion of heterogeneous data, the composability of reusable functions, and the flexibility in their cooperations. To enable such a collaboration, this paper proposes a federated platform, called Future Mobility Sensing Advisor (FMSA), which can 1) manage the three elements through standardized interfaces separately and uniformly; 2) create a fully connected knowledge graph to orchestrate the three elements efficiently and effectively; 3) support the client-server interaction in centralized and federated modes to handle service requests and edge resources with various availability and accessibilities jointly and adaptively; and 4) accommodate various mobility services to foster harmonious and sustainable mobility tenderly and invisibly. Moreover, the efficiency and effectiveness of the platform are also tested through a performance evaluation, and a pilot supported at the Great Boston Area, respectively. As a result, it shows that FMSA can 1) achieve high performance by using the two interaction modes selectively, and 2) renovate smart mobility towards sustainability through personalized services that can measure user preferences and system objectives mutually.

The bias-variance trade-off is a well-known problem in machine learning that only gets more pronounced the less available data there is. In active learning, where labeled data is scarce or difficult to obtain, neglecting this trade-off can cause inefficient and non-optimal querying, leading to unnecessary data labeling. In this paper, we focus on active learning with Gaussian Processes (GPs). For the GP, the bias-variance trade-off is made by optimization of the two hyperparameters: the length scale and noise-term. Considering that the optimal mode of the joint posterior of the hyperparameters is equivalent to the optimal bias-variance trade-off, we approximate this joint posterior and utilize it to design two new acquisition functions. The first one is a Bayesian variant of Query-by-Committee (B-QBC), and the second is an extension that explicitly minimizes the predictive variance through a Query by Mixture of Gaussian Processes (QB-MGP) formulation. Across six simulators, we empirically show that B-QBC, on average, achieves the best marginal likelihood, whereas QB-MGP achieves the best predictive performance. We show that incorporating the bias-variance trade-off in the acquisition functions mitigates unnecessary and expensive data labeling.

In this paper, a multistep sensitivity analysis (SA) approach for model calibration is proposed and applied to a complex traffic simulation model with more than 100 parameters. Throughout this paper, it is argued that the application of SA is crucial for true comprehension and the correct use of traffic simulation models, but it is also acknowledged that the main obstacle toward an extensive use of the most sophisticated techniques is the high number of model runs usually required. For this reason, we have tested the possibility of performing a multistep SA, where, at each step, model parameters are grouped on the basis of possible common features, and a final SA on the parameters pertaining to the most influential groups is then performed. The proposed methodology was applied to an urban motorway case study simulated using MITSIMLab, a complex microscopic traffic simulator. The method allowed the analysis of the role played by all parameters and by the model stochasticity itself, with 80% fewer model evaluations than the standard variance-based approach. Ten model parameters accounted for a big share in the output variance for the specific case study. A Kriging metamodel was then estimated and integrated with the multistep SA results for a global calibration framework in the presence of uncertainty. Results confirm the great potential of this approach and open up to a novel view for the calibration of a traffic simulation model.