This study proposes a real-time OD demand calibration framework that integrates data-driven origin–destination (OD) demand calibration with black-box microscopic stochastic traffic simulators. To enable bi-directional information exchange between the data-driven OD calibration and the non-differentiable traffic simulator, a metamodel-based backpropagation scheme is developed, allowing simultaneous updates of both the data-driven model and simulator parameters. In addition, the data-driven model is pre-trained offline using the metamodel to accelerate the initial calibration process. Validation through two toy networks and a Tokyo expressway corridor case study demonstrates the framework’s capability to dynamically adapt to evolving traffic conditions under various disruption scenarios. The results highlight the potential of hybrid approaches that couple data-driven and model-based techniques to support proactive traffic management, offering substantial improvements over conventional reactive systems.
Child-bicyclist crash-severity data are strongly imbalanced, making ordinary accuracy potentially misleading. This study compared three tabular deep-learning models with machine learning and statistical reference models under a train–validation–test design. The analysis used 2394 Texas child-bicyclist crashes from 2017 to 2022 classified as killed or incapacitating injury (KA), possible or non-incapacitating injury (BC), or other/no injury (O). The original records were partitioned before feature selection, preprocessing, and imbalance treatment into training (n = 1454), validation (n = 475), and untouched test (n = 465) sets. Random Forest and XGBoost rankings were calculated using the training partition only, yielding a common 10-predictor input. Validation performance was used to select class weighting for XGBoost and training-only Synthetic Minority Over-sampling Technique (SMOTEN) for the three deep-learning models. The final primary-model estimates were averaged across five seeds. Adaptive Relation-Modeling (ARM)-Net had the highest primary-model accuracy (0.662 ± 0.020) and macro-F1 (0.335 ± 0.011); XGBoost had the highest primary-model balanced accuracy (0.340 ± 0.008); and FT-Transformer had the highest primary-model macro-ROC-AUC (0.540 ± 0.006). However, the dummy-majority accuracy was 0.787 because BC dominated the test set and multinomial logistic regression exceeded all four primary models in balanced accuracy (0.378), macro ROC−AUC (0.561), and macro average precision (0.361). Minority-class discrimination was weak, bootstrap intervals overlapped substantially, and SHapley Additive exPlanations (SHAP) rankings showed limited stability relative to cross-seed permutation importance. Class-balanced performance was limited across all models. The evaluated models should be treated as transparent benchmarks rather than operational prediction tools; larger, externally validated datasets with richer exposure and behavioral information are needed. From an AI-for-transportation-safety perspective, the study contributes a leakage-controlled benchmarking protocol and an explanation-stability check, showing that advanced tabular deep learning models should be judged against simple baselines and class-balanced metrics before their outputs are used to support safety decisions.
This paper studies the behaviors of Level-4 autonomous vehicles based on the large-scale data of Waymo self-driving cars (SDCs). Through a careful and rigorous treatment of the data, we have identified the stationary traffic states and used them to obtain the spacing-speed relationship, which is then used to derive the flow-density relationship. We have analyzed the Waymo SDCs and human-driven vehicles (HDVs) in the traffic stream. It was found that the SDCs maintain a linear spacing-speed relationship, like their human counterparts, but the SDCs are more conservative (characterized by a larger spacing), which results in a significant downgrade of roadway capacity ( about 18
When combining telematics data with infrastructure and road network data, new opportunities for spatial road safety analysis are opened, this may although introduce challenges. This study proposes an improved method for aggregating smartphone-based telematics data into entities, such as nodes and edges, to support more robust spatial road safety analysis. A telematics dataset comprising multiple trips represented by geospatial coordinates was aggregated to edges using a nearest-edge approach and to nodes using a buffer-based method, introducing inconsistencies. A refined node aggregation method was introduced, restricting data points to edges directly connected to the buffer’s origin node, resulting in node-level variability reductions of approximately 7–50
The increase in the need for sustainable urban mobility has made carsharing systems an interesting alternative to personal vehicle ownership. This study proposes a methodology allowing the expansion of carsharing systems by choosing optimal station locations using MATSim simulations together with advanced optimization algorithms. Using an accessibility metric, this paper identifies optimal station locations that maximize user accessibility. The methodology was applied to Montreal’s carsharing network to demonstrate its benefits. The expansion of carsharing system process used iterative simulations that assess the baseline and extended scenarios. To identified target regions and ideal station placements, a clustering method and TABU search algorithms were used. Significant improvements in trip counts and distances, particularly during peak hours, are demonstrated by the outcomes of ten expansion cycles. The suggested method demonstrated more alignment with user behavior and geographical demand, enhancing overall system usage, in contrast to selecting random carsharing station sites within the targeted areas. All things considered, this study shows how simulation and optimization methods can be combined to tackle the challenges of carsharing network planning. In addition to the methodological contribution, this study highlights how simulation and optimization outcomes can be used to develop urban policies and supporting planners and carsharing operators in designing stations’ deployment strategies. To further improve demand projections and make sure that variations in urban settings are taken into account, future study should add dynamic aspects, such as weather and temporal variations.
