Traffic prediction based on real-world traffic data is a crucial task in Intelligent Transportation Systems (ITS). However, the issue of missing observations due to real-world disturbances undermines the robustness and accuracy of traffic prediction. This problem necessitates the development of a prediction model that integrates the imputation mechanism to be compatible with missing observations. This paper introduces ATTST, a self-imputation-assisted prediction model specifically designed to address the challenge of missing observations in the traffic prediction task. Unlike traditional approaches utilizing an additional supervised imputation model before prediction, the imputation unit in our model does not need the extra label for the missing observations. Our model employs a self-imputation unit to impute the missing observations by partially masking the observed data as the ground true labels. Thus, the self-imputation unit along with an encoder–decoder architecture and a graph evolving unit together directly predict future traffic data with multi-level missing observations. The effectiveness of ATTST is validated using several real-world traffic datasets, including speed and flow data, across various multi-step prediction scenarios with diverse missing observations. These validations demonstrate the model’s robustness and practical applicability in real-world traffic prediction tasks. The results show that ATTST can reliably predict traffic conditions even with incomplete data, making it a valuable tool for traffic management and planning. • The problem of imputation and prediction for traffic speed and flow data with missing observations is formulated within an end-to-end framework. • The proposed AttSt model addresses this problem by incorporating a self-imputation mechanism to effectively handle missing observations. • The model employs a graph network to capture and process spatiotemporal correlations within the traffic data. • Comprehensive evaluations on four real-world traffic speed and flow datasets validate the effectiveness of the proposed approach.

Connected vehicle (CV) data are increasingly available and widely used in transportation engineering for traffic monitoring, safety analysis, and infrastructure planning. However, the representativeness of CV data in general urban mobility analysis remains underexplored, raising concerns about potential biases between observed mobility patterns in CV data and actual travel behaviors, particularly across different demographic and socioeconomic groups. This study compares Wejo connected vehicle (CV) data with SafeGraph mobile phone records covering boarder population to reveal the representativeness of connected vehicle data in the context of urban mobility analysis. Using entropy-based measures of destination income diversity and normalized inter-neighborhood interaction strength, we examine how each dataset reflects mobility structures across income groups in San Antonio, Texas. Results show that SafeGraph data capture more multimodal and socially integrative travel behaviors, particularly among low-income communities, while Wejo data primarily reflect routine, vehicle-based movements concentrated among higher-income users. Interaction patterns in the CV data are more spatially clustered, with stronger flows observed within affluent neighborhoods. These differences underscore the behavioral and demographic selectivity embedded in CV data and point to important limitations when using such data to analyze mobility-based segregation. The findings contribute to ongoing efforts to evaluate the representativeness of emerging mobility datasets and their implications for urban spatial analysis.

Traditional urban planning frequently overlooks Children’s Independent Mobility (CIM) – the developmental stage where children begin to navigate their environment without adult supervision. Current educational methods often fail to bridge the ‘knowledge-behaviour gap’ which represents the disconnect between knowing traffic safety rules and applying them in real-world situations. To address this, this study presents a site-specific geogame using a Minecraft digital twin of two school neighbourhoods in Utrecht, Netherlands. Both the game environment and the engagement process were co-designed with traffic-focused local authorities to ensure relevance to real issues. Primary school children navigated the virtual environment and responded to embedded spatial questionnaires regarding traffic safety rules around their schools. The results demonstrate that this approach shifts children’s roles from passive learners to active ‘citizen-experts’. Analysis of in-game movement and chat logs revealed specific spatial insights actionable for planners, including the identification of perceived danger zones and preferred shortcuts. The study concludes that geogames may offer a scalable method to capture children’s situated spatial knowledge and effectively integrate their perspectives into the urban design process.

Urban infrastructure works conducted under live traffic conditions often face a persistent gap between approved traffic management plans and their actual field implementation. This gap remains underexplored in longitudinal studies, particularly in utility projects from low- and middle-income urban contexts. This study evaluates a risk-based supervisory approach that integrates daily monitoring of the Traffic Management Plan (TMP) with a corporate risk management framework aligned with ISO 31000. The dataset includes 288 supervised workdays over 16 months (November 2023–February 2025), 99 non-conformity tickets, 96 signal-theft events (137 units), and seven traffic incidents. The analysis combines descriptive statistics, hypothesis testing, logistic regression, segmented longitudinal analysis, count models, response-time evaluation, and a composite risk index. TMP non-compliance decreased from 18.8% to 6.9% between the first and second halves of the study period (p=0.0028). The odds of non-compliance were significantly higher during the staff transition period in April–May 2024 (OR = 3.50; 95% CI: 1.24–9.82), while day and night shifts showed comparable rates. Monthly patterns indicate that staff instability and signal theft contributed to non-compliance levels, and ticket resolution remained slow (mean response time: 69.9 days). These findings highlight the importance of supervisory continuity, contractor stability, and timely corrective actions in improving work zone safety.

This study investigates the critical challenges associated with location selection for large-scale Agricultural Product Wholesale Markets (APWMs) under the traditional circulation model. It identifies and elaborates on the evolving characteristics of circulation stakeholders, supply chains, distribution channels, organizational structures, and external environments during the transition from traditional to modern circulation systems. In response to the demands of modern circulation, a comprehensive location selection evaluation framework is proposed, integrating five key criteria: location, planning, transportation, land use, and urban compatibility. Elastic and rigid evaluation standards are established according to the nature of each criterion. The framework innovatively integrates national territorial and spatial planning, road traffic planning, industrial development planning, and urban big data resources through Geographic Information System (GIS) technology, consolidating these into a unified database. To determine comprehensive weights for different functional types of APWMs, the normalized linear aggregation method is applied to combine weights derived from the Analytic Hierarchy Process (AHP) and the entropy weight method, enabling an analysis of correlation and contribution levels. Furthermore, this study introduces an innovative application of the Genetic Algorithm (GA), implemented in Python, to re-optimize the integration of subjective and objective weights through iterative computation until convergence, thereby enhancing the accuracy of comprehensive weight estimation and validating the location selection outcomes. A case study demonstrates the successful development of a five-phase location selection methodology—region screening, scope delineation, data analysis, weight optimization, and comprehensive evaluation—enabling both quantitative ranking and recommendation of candidate location and the optimal solution was ultimately selected from seven candidate schemes. This research provides practical guidance for location selection of large-scale APWMs within modern circulation contexts and offers methodological insights applicable to urban logistics planning and the siting of other large-scale infrastructure facilities.

Effective monitoring of unusual transportation behaviors, such as wrong-way cycling (i.e., riding a bicycle or e-bike against designated traffic flow), is crucial for optimizing law enforcement deployment and traffic planning. However, accurately recording all wrong-way cycling events is both unnecessary and infeasible in resource-constrained environments, as it requires high-resolution cameras for evidence collection and event detection. To address this challenge, we propose WWC-Predictor, a novel method for efficiently estimating the wrong-way cycling ratio, defined as the proportion of wrong-way cycling events relative to the total number of cycling movements over a given time period. The core innovation of our method lies in accurately detecting wrong-way cycling events in sparsely sampled frames using a light-weight detector, then estimating the overall ratio using an autoregressive moving average model. To evaluate the effectiveness of our method, we construct a benchmark dataset consisting of 35 minutes of video sequences with minute-level annotations. Our method achieves an average error rate of a mere 1.475% while consuming only 19.12% GPU time required by conventional tracking methods, validating its effectiveness in estimating the wrong-way cycling ratio. Our source code is publicly available at: <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/VICA-Lab-HKUST-GZ/WWC-Predictor</uri>

Traffic flow prediction is vital in urban traffic management, planning, and development. With the continuous advancement of urbanization, there is an increasing demand for traffic flow prediction models to achieve higher accuracy and long-range forecasting capabilities. Against this backdrop, traditional methods that rely on local feature extraction and static spatial graph construction often fall short of expectations. This highlights the urgent need for advanced approaches to dynamically model spatio-temporal features while capturing global dependencies, effectively meeting the demands of complex traffic flow prediction tasks. To achieve this, we propose the Transformer-Enhanced Adaptive Graph Convolutional Network (T-AGCN), a novel model designed to capture global temporal relationships and dynamically extract rich spatial information. T-AGCN incorporates an Adaptive Graph Learner module to model dynamic relationships among traffic nodes and a Transformer-Based Spatio-Temporal graph convolutional module to capture long-range temporal dependencies in historical traffic data effectively. These innovations enable T-AGCN to jointly learn dynamic spatial interactions and complex temporal patterns, offering a comprehensive representation of traffic network dynamics. We evaluate T-AGCN on three real-world datasets, PeMSD7(M), PeMS08, and METR-LA. The experimental results demonstrate that T-AGCN, inspired by the baseline model Spatial-Temporal Graph Convolutional Network (STGCN), significantly enhances its design. Moreover, T-AGCN consistently outperforms state-of-the-art models, including the Transformer-Based Interactive Temporal and Adaptive Network (TITAN) and the Spatial-Temporal Decoupled Masked Autoencoder (STD-MAE). The implementation is available on GitHub at https://github.com/time1722/T-AGCN .

Origin-Destination (OD) flow matrices are critical for urban mobility analysis, supporting traffic forecasting, infrastructure planning, and policy design. Existing methods face two key limitations: (1) reliance on costly auxiliary features (e.g., Points of Interest, socioeconomic statistics) with limited spatial coverage, and (2) fragility to spatial topology changes, where reordering urban regions disrupts the structural coherence of generated flows. We propose Sat2Flow, a structure-aware diffusion framework that generates structurally coherent OD flows using only satellite imagery. Our approach employs a multi-kernel encoder to capture diverse regional interactions and a permutation-aware diffusion process that maintains consistency across regional orderings. Through joint contrastive training linking satellite features with OD patterns and equivariant diffusion training enforcing structural invariance, Sat2Flow ensures topological robustness under arbitrary regional reindexing. Experiments on real-world datasets show that Sat2Flow outperforms physics-based and data-driven baselines in accuracy while preserving flow distributions and spatial structures under index permutations. Sat2Flow offers a globally scalable solution for OD flow generation in data-scarce environments, eliminating region-specific auxiliary data dependencies while maintaining structural robustness for reliable mobility modeling.

Traffic flow prediction is a crucial aspect of transportation planning to prevent congestion. A good understanding of a traffic planning strategy will assist in avoiding traffic congestion. Complex factors, such as inter-region traffic conditions, vehicle relationships, and unexpected incidents, influence accurate traffic flow prediction. Thanks to embedded sensors in transportation systems, vehicle movement data can be widely collected and analyzed to predict the traffic flow. To this end, this study proposes a deep learning-based prediction algorithm called Deep learning-based traffic flow graph prediction (DeepTFGP) for forecasting traffic flow on each road within the road network. We modeled road networks and traffic flow data using the graph theory to generate a traffic flow graph (TFG) representing the diversion of vehicle flow in the traffic network. Through learning from the TFG dataset, DeepTFGP accurately predicts future traffic flow. DeepTFGP employs three independent residual neural network modules to model the temporal characteristics of traffic flow: closeness, period, and trend. Each module consists of multiple residual units, and the outputs are combined through a shared time prediction function module. The DeepTFGP parameters are automatically optimized using the gradient descent algorithm of neural networks to achieve accurate traffic flow prediction. We collected and constructed a relevant UK National Highways dataset from traffic detection data provided by the British Department of Transport on the UK highway network. We performed experiments and comparisons on this dataset with existing algorithms. Our final results demonstrate that DeepTFGP outperforms various benchmark comparison models in accuracy.

Amid the rapid development of the “Internet Plus” sharing economy, shared mobility grapples with challenges like imprecise ride path detection and unreasonable speed settings. Using over 280,000 September 2025 Citi Bike ride records in New York City, this study establishes a three-dimensional stacked recognition model (integrating hotspot identification, vectorization, terrain modeling and network analysis) via MiniBatchKMeans clustering and other terrain feature techniques to identify major cycling paths and determine scientific e-bike speeds. The model locates key terrain areas, constructs a main cycling network covering over 60% of core traffic, and reveals NYC cycling’s multi-centered, corridor-based pattern linked to transport and commercial zones; it further proposes a zoned dynamic speed limit scheme (15 km/h in rush hours/ central areas, 20 km/h on main routes, 25 km/h in valleys/other regions), providing data and methodological support for optimizing urban slow traffic planning, regulating shared mobility and promoting coordinated integration with urban transport systems.

Traffic accident detection and object detection have become key areas of research due to their direct impact on safety, traffic congestion mitigation, and intelligent traffic planning. This study presents a structured analysis of classical detection methods and artificial intelligence-based techniques, highlighting their methodologies, objectives, and performance results. The study categorizes existing research into threshold-based approaches, statistical approaches, image processing, rule-based approaches, and machine learning approaches, with further emphasis on predictive modeling, graph-based approaches, and optimization approaches. Considerable emphasis is placed on identifying systems that are capable of operating under adverse weather conditions such as fog, rain, and snow. These scenarios significantly affect detection accuracy. Although several authors incorporate environmental resilience into their models, most studies still evaluate performance under ideal conditions, revealing a critical gap in research. This analysis highlights the need to develop robust detection mechanisms that can adapt to real-world variability and environmental disturbances. Findings show that AI-based methods significantly outperform classical approaches in terms of adaptability and scalability, but their dependence on training data limits their performance in adverse conditions. The study concludes with recommendations for future work to prioritize multimodal sensing, generalization across weather conditions, and integration of environmental intelligence to ensure reliable real-time detection of traffic events under all operating conditions.

Traffic congestion is a critical challenge in urban transportation systems leading to longer travel times, higher fuel consumption, and increased environmental pollution. This study examines traffic congestion along the major urban road corridor in Thiruvananthapuram, Kerala, India, by Geographic Information Systems (GIS). Travel time data were collected from Google Maps under varying conditions, including weekdays, weekends, holidays, festivals, in different weather conditions, and special events, to capture temporal and situational variations in traffic flow. The collected data were analyzed to identify congestion zones based on speed across different road segments. QGIS software was utilized to perform Optimal Route Analysis by Contraction Hierarchies (CH) within the study area using the OpenRouteService (ORS) Directions API with the driving car profile, based on OpenStreetMap road network data. The study demonstrates that GIS provides an effective framework for traffic congestion analysis, route optimization, supporting improved urban traffic management and planning.

Mid-sized Philippine cities commonly rely on fixed-time traffic signal plans that cannot respond to short-term, demand-driven surges, resulting in measurable idle time at stop lines, increased delay, and unnecessary emissions, while adaptive signal control has demonstrated performance benefits, many existing solutions depend on centralized infrastructure and high-bandwidth connectivity, limiting their applicability for resource-constrained local government units (LGUs). This study reports a field deployment of TrafficEZ, a lightweight edge AI signal controller that reallocates green splits locally using traffic-density approximations derived from cabinet-mounted cameras. The controller follows a macroscopic, cycle-level control abstraction consistent with Transportation System Models (TSMs) and does not rely on stationary flow–density–speed (fundamental diagram) assumptions. The system estimates queued demand and discharge efficiency on-device and updates green time each cycle without altering cycle length, intergreen intervals, or pedestrian safety timings. A quasi-experimental pre–post evaluation was conducted at three signalized intersections in El Salvador City using an existing 125 s, three-phase fixed-time plan as the baseline. Observed field results show average per-vehicle delay reductions of 18–32%, with reclaimed effective green translating into approximately 50–200 additional vehicles per hour served at the busiest approaches. Box-occupancy durations shortened, indicating reduced spillback risk, while conservative idle-time estimates imply corresponding CO2 savings during peak periods. Because all decisions run locally within the signal cabinet, operation remained robust during backhaul interruptions and supported incremental, intersection-by-intersection deployment; per-cycle actions were logged to support auditability and governance reporting. These findings demonstrate that density-driven edge AI can deliver practical mobility, reliability, and sustainability gains for LGUs while supporting evidence-based governance and performance reporting.

This paper proposes a novel Physics-Constrained Neural Network (PCNN) approach for solving the Dynamic Origin-Destination Estimation (DODE) problem. The method combines physical constraints with deep learning, utilising traffic flow observations and network structure to accurately estimate dynamic OD matrices. This study introduces a novel spatiotemporal structure (CNN & Line-based Network Graph Aggregation) for high-accuracy estimation. By employing a Physics-Constrained Deep Learning (PCDL) and Dynamic Traffic Assignment Learner (DTAL) framework, the model integrates a machine learning surrogate into the model-driven component. This approach avoids encoding complex terms directly in the DTA process, thereby improving accuracy. Comprehensive evaluations on two networks validate the method's superior accuracy and data efficiency compared to benchmarks. Experimental results demonstrate the PCNN's robustness under various demand levels and missing data scenarios, providing a significant novel approach for traffic planning and management.

Accurate prediction of sensor network data in critical domains such as electric power systems and traffic planning is a core task for ensuring grid stability and enhancing urban operational efficiency. Although deep learning models have achieved significant architectural advancements, their training strategy implicitly assumes that all future events are equally predictable, ignoring that the future evolution of sensor signals intertwines deterministic patterns with stochastic events and that prediction difficulty increases with temporal distance. Forcing a model to fit inherently unpredictable events with a uniform supervision may impair its ability to learn generalizable patterns. To address this, we introduce an Unpredictability Perception loss that dynamically computes a supervision weight. The computation of this weight unifies two assessment dimensions of the intrinsic unpredictability of the forecasting task. The first originates from a posterior analysis of the signal content’s randomness, while the second stems from an a priori consideration of temporal distance. The first dimension, through a complexity-aware weight derived from local spectral entropy, reduces supervision on random segments of the signal. The second dimension, through a temporal decay weight based on exponential decay, lessens supervision for distant future points. Applied to the advanced TimeMixer model, experimental results show that our approach achieves performance improvements across multiple public benchmark datasets. By matching the supervision strength to the intrinsic predictability of the signals, our proposed Unpredictability Perception loss function enhances the forecasting accuracy for sensor network data, providing a more reliable technical foundation for ensuring the stability of critical infrastructures like power grids and optimizing urban traffic systems.

Traffic flow prediction is of great importance in traffic planning, road resource management, and congestion mitigation. However, existing prediction have significant limitations in modeling multi-scale spatial-temporal features, particularly in capturing temporal periodicity and spatial dependency in dynamically evolving traffic networks. This paper proposes a novel framework of traffic flow prediction, referred to as Adaptive Graph Fusion Dual-scale Convolutional Network (AGFDCN), which integrates spatial-temporal dynamic graphs with dual-scale convolutional networks. Specifically, we introduce a Dual-Scale Temporal Network, which combines long- and short-term dilated causal convolutions with a temporal decay-aware attention mechanism to efficiently capture traffic patterns across multiple temporal scales. Furthermore, we design a Dynamic Adaptive Graph Module, which models complex spatial dependencies in traffic networks through an adaptive graph fusion mechanism and a dual-path attention-gated module. Finally, the temporal and spatial representations are integrated by employing a gated fusion mechanism, enhancing the overall prediction performance. Experimental results obtained based on three highway datasets (i.e., PEMS04, PEMS07 and PEMS08) verify that the proposed model outperforms several state-of-the-art baselines in various evaluation metrics. Compared to the spatial-temporal graph model AGCRN with best performance in the baseline models, the proposed model exhibits significant improvements across all datasets: it achieves reduces of MAE by 42.07% and RMSE by 35.43% on PEMS04; MAE by 28.35% and RMSE by 29.28% on PEMS07; and MAE by 30.52% and RMSE by 30.73% on PEMS08, respectively, validating its effectiveness in modeling complex spatial-temporal traffic data and its robustness in handling sudden traffic changes.

An accurate forecast of general aviation traffic is crucial for air navigation service providers, as it affects the overall efficiency of air traffic management and capacity planning. This paper presents a deep learning methodology for predicting general aviation traffic, combining calendar and meteorological sources through a detailed feature-engineering procedure. The approach is rigorously evaluated using historical data from the Nice Cote D’Azur Terminal Control Center sectors, resulting in a significant 32% enhancement in global prediction performance with recurrent neural network models compared to existing operational tools. The paper explores additional analysis techniques to gain a deeper understanding of the predictions that are produced by each model.

Microscopic traffic dynamics deduction, capable of reconstructing and predicting traffic states at 0.1–0.5-s temporal and 0.1–1-m spatial granularity with limited information, constitutes a critical foundation for traffic control and planning. Conventional deduction methods based on analytical car following models are oversimplified in parameters, and thus, they inadequately capture complex traffic dynamics, while higher-dimensional parametric deep learning-based methods suffer from poor iteration speed. As a compromise between speed and modeling capability, this article proposes a scene-to-scene autoregressive framework that fuses simulation, data-driven prediction, and real-time monitoring for microscopic traffic dynamics deduction. The methodology implements scene-level autoregressive deduction cycles rather than agent-based modeling while enhancing real-time accuracy through integration of connected vehicle (CV) future information. Evaluations using the International, Adversarial, and Cooperative Motion dataset demonstrate that the proposed simulation framework achieves comparable accuracy in a reasonable runtime. Real-time CV information incorporation is also proved to be effective for further accuracy improvements.

Transportation agencies manage hundreds of active work zones, often spread across large geographic regions. Traditionally, verifying compliance with maintenance of traffic (MOT) plans has required in-person inspections by staff or contractors which is time-consuming, resource-intensive, and often requires extensive driving with an image or video recording device. With tight staffing requirements, it is only possible to inspect a small subset of work zones using field visits. The emergence of commercial truck dash cameras that can provide images at 1 second intervals from several trucks a day now enables an agency to review dash camera images to “virtually” drive their work zones. This has the benefit of not only drastically reducing travel time and costs, but also provides an opportunity to perform repetitive, weekly inspections of the work zones to determine compliance with an agency’s practices. A series of case studies covering sign placement, lane configuration, temporary work zones, maintenance closures, and pavement marking applications is presented. The paper concludes with an example of two different artificial intelligence (AI) models that can be used to process these images to determine the presence of a work zone on interstate roads. Across a suite of 40 images from 8 states and 20 interstate routes, both models performed very well and demonstrate considerable opportunity to integrate commercial dash camera images with AI models to screen work zones at scale for further human review.

The Origin–Destination (OD) matrix is essential for traffic planning as well as for the management of transportation logistics and operations. It represents the movement patterns of individuals and vehicles between different locations. By understanding how traffic flows can interact with the road network, transportation authorities can implement proactive safety measures to reduce accident risks. Effective accident management strategies also rely on knowledge of traffic flow patterns, which the OD matrix provides. However, a sufficiently large dataset is required to accurately estimate an Origin–Destination matrix from traffic counts and to ensure that the resulting predictions are broadly applicable. Due to the lack of publicly available datasets in the area of our investigation, we leveraged the Simulation of Urban Mobility (SUMO), a widely used traffic simulator in the transportation domain, to generate a new dataset and estimate OD matrices from synthetic traffic counts. Our study includes DFRouter, a simulation-based tool, not as a direct benchmark but rather as a reference point to illustrate the effectiveness of machine learning (ML) models in OD matrix estimation. In a comparative analysis, we trained and tested various ML models, including an Artificial Neural Network (ANN) for deep feature extraction combined with a Support Vector Regression (SVR) model in a hybrid ANN+SVR approach. Overall, all ML models employed in our study performed significantly better than DFRouter. The hybrid ANN+SVR model achieved the lowest error rates on our dataset across different OD configuration scales.