Abstract This study proposes a minute-level tidal lane control framework for highway toll plazas operating under strongly time-varying and directionally imbalanced demand. Historical and real-time traffic volumes, lane-specific service rates, holiday indicators, and weather variables are used to generate 30-min-ahead directional flow forecasts with a Transformer–bidirectional long short-term memory model. The forecasts are then combined with overload-risk screening and a directional-imbalance measure based on the volume-to-capacity ratio to determine whether lane reversal should be implemented. Using field data from a major highway toll plaza in China, the proposed intelligent tidal lane control method reduces average travel time by about 20% during the morning peak and 22% during the evening peak relative to the traditional reactive strategy, with larger gains under high-demand tidal-surge conditions and limited unnecessary intervention under low-demand conditions. The study also provides calibration guidance and recommended ranges for the key thresholds α and d, offering a transferable parameter-calibration procedure and practical guidance for real-world toll-plaza operations.
Abstract Electric bike sharing (EBS), a convenient and efficient mobility solution, has become an integral part of urban transportation systems in many cities. To promote the sustainable development of EBS systems, it is vital to understand the spatiotemporal patterns of EBS demand and their relationships with influencing factors. Using 1 month of EBS trip data from Yangzhou, China, this study applied a nonnegative matrix factorization (NMF) model, combined with the space-time cube model and emerging hot spot analysis, to identify spatiotemporal patterns of EBS demand. The results show that EBS demand exhibits four distinct temporal patterns: (1) morning peak, (2) daytime, (3) evening peak, and (4) nighttime, each with unique spatial hot spot distributions. Subsequently, the study employed Light Gradient Boosting Machine (LightGBM) models to explore the complex relationships between EBS demand and the built environment within each of these patterns, with a pooled model as a benchmark. The analysis reveals that the relative importance of variables varies significantly across patterns. For instance, the importance of the percentage of residential district peaks in the morning peak pattern, while the percentage of commercial district and educational institutions are more prominent in the daytime, evening peak, and nighttime patterns. Moreover, these relationships exhibit pronounced nonlinearity. For example, the positive correlation between population density and EBS demand levels off at 2,000 people / km 2 for the morning peak pattern, but not until 5,000 people / km 2 for other patterns. These findings can guide operators in formulating refined operational strategies and assist policymakers in developing targeted interventions.
Abstract Optimized geometric design of a tunnel is crucial for enhancing its operational efficiency as well as assessing traffic safety. This study utilizes a continuous prediction model employing a three-segment feature aggregation method that considers upstream, current, and downstream segments to estimate driver behavior at any position within the tunnel. The geometric features of different positions, ranges, and quantities were compared to obtain appropriate geometric features that reduce the features’ deviation from the driver’s perception. Through orthogonal experimental design, training road sections were developed incorporating horizontal and vertical alignment combinations. An integrated extreme gradient boosting–Shapley additive explanation method (XGBoost-SHAP) model was proposed to predict operational speed, acceleration, and trajectory deviation at arbitrary segments in the tunnel. The proposed three-segment prediction model outperforms traditional models considering only upstream and downstream segment, achieving higher prediction accuracy. Compared to conventional approaches, the XGBoost-SHAP model reduced the root mean square error by 8% on the validation tunnel. Key factors are identified influencing vehicle operational speed include the average curvature from 400 to 200 m upstream, the average absolute curvature and the grades from 200 m upstream to the current segment, and the average curvature and average grade from the current segment to the farthest visible downstream. Results indicate that vehicle operational speed is mainly influenced by the gradient of the current segment. Vehicle acceleration rate is influenced by the gradient changes in current segment. Vehicle trajectory deviation is impacted by curvature in the current and downstream segments. Recommendations are further proposed for geometric design of expressway tunnels. These insights are pivotal for evaluating and enhancing tunnel designs to optimize traffic efficiency.
Abstract Compared with conventional evaluations that focus on bus service demand, bus service effectiveness (SEV) emphasizes both the demand and supply dimensions of service. However, knowledge on spatial cost-effectiveness analyses of SEV and its relationship with spatial factors such as the built environment remains limited. Using smart card data from Shanghai, this study investigates how the built environment affects bus SEV using a slack-based efficiency measure and machine learning models. The results reveal that less bus ridership does not necessarily equate to lower bus SEV. Population and employment density show distinct effects on bus SEV, with employment density exhibiting a threshold effect at 5,000 jobs / km 2 . These findings offer insights for planners to design targeted built environment interventions that enhance SEV and underscore the importance of adopting a more spatially cost-effective evaluation framework that goes beyond bus ridership alone.
Abstract Transportation planning requires anticipating how built environments, population characteristics, and aggregate travel indicators evolve over time. Conventional forecasting approaches often treat these dimensions separately, limiting their ability to represent spatial-demographic change. This study introduces the Demographics-Informed Neural Network (DINN), a deep learning framework that forecasts next-year county-level satellite imagery from historical image sequences while enforcing demographic consistency through a frozen demographic predictor. DINN combines an encoder-decoder image predictor, gated skip connections, demographic bottleneck fusion, and demographic consistency regularization. The frozen demographic encoder is then reused as a fixed feature extractor to estimate aggregate county-level commute-mode and vehicle-availability indicators. Experiments using US county data from 2012 to 2023 show that DINN achieves an MSE of 0.00379 and SSIM of 0.8342 for satellite image prediction. Across repeated grouped spatial splits, DINN provides modest but consistent improvements over ConvLSTM, transformer-based, and encoder-decoder baselines, with stronger performance on both image reconstruction and change-detection metrics. The demographic predictor explains approximately 80% of the variance in population characteristics from satellite imagery. Transfer learning results show higher accuracy for absolute travel-indicator counts than for scale-controlled shares, indicating that part of the signal reflects county population scale. Overall, the findings suggest that satellite-derived spatial representations capture built-environment and demographic correlates of aggregate travel indicators; however, the results should be interpreted as county-level predictive associations rather than causal behavioral inference.
