Journal of Advanced Transportation最新文献

The increasing scale of highway reconstruction and expansion projects has intensified traffic management challenges in construction zones, particularly within sparse road networks constrained by limited diversion capacities and elevated freight truck ratios. This study proposes an integrated analytical framework that combines microscopic simulation, automated machine learning (AutoML), and explainable artificial intelligence. Traffic flow dynamics under high truck proportions (72%) were modeled using the VISSIM microsimulation, generating 1320 parameterized scenarios encompassing traffic volume, work zone length, speed limits, and vehicle composition. By leveraging the AutoGluon AutoML framework, we developed an ensemble delay prediction model using optimized feature engineering. SHapley Additive exPlanations (SHAP) interpretability analysis further decoded the multifactorial coupling mechanisms influencing traffic organization. The results demonstrate that while complex ensembles achieved the lowest error (RMSE = 1.49), the CatBoost_BAG_L1 model was identified as the optimal model for operational deployment, achieving identical accuracy with a more than 25-fold improvement in computational speed. The SHAP-based interpretation revealed traffic volume as the dominant delay contributor, exhibiting nonlinear dynamics with escalating marginal effects beyond 1400 vehicles/h. Increasing the speed limit to 80 km/h elevated delays by 0.58 units, while work zones exceeding 2 km in length induced length-proportional delay amplification. This methodology advances intelligent decision-making for dynamic lane control and truck scheduling optimization in diversion-constrained environments.

The purpose of this study was to explore the risk factors for autonomous vehicle (AV) crashes and their interdependencies. A total of 659 AV crash data were collected between 2018 and July 2024 from AV crash reports published by the California Department of Motor Vehicles. Characteristics such as crash location and time, driving patterns, vehicle motion, crash type and vehicle damage, and traffic conditions were considered as potential risk factors in the study. Considering the multilevel and multidimensional nature of the crash data, the study adopted an association rule mining (ARM) approach to identify the risk factors that frequently occur in AV crashes. By improving the Apriori algorithm, based on the traditional Apriori algorithm, the association rule judgment index is added, and the accuracy and mining efficiency of association rules are improved. The results show that rear-end collisions in autonomous driving mode are more serious, especially when stopping at intersections, while the rear vehicle chooses to continue driving or slow down. Accident risk is higher at night, with on-street parking and 2-lane conditions in both directions. The occurrence of no-damage and minor crashes is more likely to be influenced by roadway characteristics and traffic conditions, and nonmotorized lanes, on-street parking, and median strips on the roadway play a key role in reducing crash damage. The results of the study inform relevant policies to improve road safety and the efficiency of AVs to enhance the overall safety of road traffic.

Driver deviations from planned routes during navigation threaten road safety and reduce traffic efficiency, with F-shaped intersections emerging as a high-risk scenario. This study investigates deviation causes using real-world navigation operation data and hourly aggregated observations. A generalized structural equation model (SEM) with a zero-inflated negative binomial link is applied to disentangle direct effects of external conditions and indirect effects mediated by traffic flow and congestion. Key findings include the following: exit/entrance roads (β = 0.135, p < 0.001) and road type (β = 0.100, p < 0.001) exert the strongest direct effects on deviations, while traffic congestion mediates 12% of the indirect effect of weather conditions on deviations. An actionable design takeaway is that extending advance signage on high-risk segments reduces deviations by 18% [−14%, −22%] at median traffic flow, providing targeted technical support for F-shaped intersection optimization to improve road safety and traffic efficiency.

To improve lane-changing efficiency and reduce safety risks for ramp vehicles in highway merging areas, this paper presents a method for predicting vehicle trajectories in these types of scenarios, and it is based on physics-enhanced residual learning. Focusing on ramp lanes and adjacent mainline lanes, the model considers the influence of both the current and target lanes on the vehicle’s velocity during lane-changing maneuvers. A hybrid prediction model is constructed by integrating a physics-based model with a data-driven approach. Specifically, an improved speed prediction model based on the Gipps general collision avoidance algorithm is introduced to calculate vehicle speed variations during lane-changing maneuvers, and its parameters are calibrated using a genetic algorithm. The next trajectory point of the vehicle is predicted, and the corresponding residual is computed using the calibrated physical model. A long short-term memory network is constructed to learn and predict the residuals. The final trajectory prediction is obtained by combining the physical model’s output with the predicted residuals. The experimental results based on real-world traffic data show that the approach introduced in this study outperforms traditional neural network models significantly in terms of both accuracy and stability. The model achieves a higher determination coefficient and notably reduces both overall and longitudinal prediction errors. Additionally, ablation studies confirm that incorporating a Gipps-based residual learning mechanism into the data-driven model significantly enhances prediction performance, thereby validating the effectiveness of integrating physical information with residual learning. The proposed trajectory prediction model offers a novel and effective solution for improving trajectory prediction accuracy for ramp vehicles in highway merging areas.

The assessment of the risk associated with urban road networks, particularly in the context of earthquakes, is of paramount importance for the identification and reinforcement of the most vulnerable sections of urban road networks and the selection of optimal emergency rescue routes. This paper proposes an innovative method combining Bayesian networks and multifactor decision theory. It considers the effect of uncertainty brought about by earthquakes on the composition of road networks and distinguishes critical sections that make up road networks under the influence of multiple factors. This enables suggestions to be made for postearthquake emergency rescue work. Seismic hazards can cause structural damage to urban road networks and affect normal access. This paper quantifies the risk of earthquakes to urban road networks and evaluates the seismic capacity of the network by estimating the travel time of each road section and the connectivity of road sections. A Bayesian network model is established, with the pre-earthquake connectivity of each partial road section defined as the priori probability. The data of the Bayesian network is updated based on the information obtained from the observation of the components and the system. Multiattribute decision theory is employed to calculate the a posteriori probability, which is then related to the travel time and the length of the road sections, as well as the critical parts of the road network and the optimal emergency paths. This paper presents a case study in which the effectiveness of the decision analysis model is verified. The results of this study contribute to the improvement of emergency rescue operations following an earthquake and reinforce critical sections of the road network in advance, thereby enhancing the overall seismic resilience of the road network.

Since 2015, China’s railway freight volume has steadily increased; however, the transportation capacity of the existing road network cannot fully adapt to the pressure caused by the increase in freight volume. In addition, the limitations of the current traffic flow adjustment system pose a significant risk of potential capacity bottlenecks in the road network. Thus, there is an urgent need to identify and eliminate dynamic capacity bottlenecks. Therefore, this paper proposes an optimization method for alleviating operational bottlenecks based on multiple strategies. First, evaluation indicators for dynamic railway transportation capacity bottlenecks are selected, and an optimization model for transportation capacity bottlenecks is designed considering the dynamic transportation capacity bottleneck evaluation indicators as constraints. Then, an improved adaptive large-scale neighborhood search algorithm is studied to solve the above model, and the existing A^∗ algorithm is improved to solve the spatiotemporal route, further enhancing the solving efficiency of the model. Finally, a numerical example is designed to verify the feasibility and effectiveness of the comprehensive railway transportation capacity bottleneck relief method designed in this paper.

The transportation sector is a major contributor to global carbon emissions; however, the high costs of infrastructure and equipment present significant barriers to effective emission reduction. As a market-based mechanism, carbon trading improves the efficiency of emission reductions and facilitates the integration of the transportation sector into the global carbon mitigation framework. Incorporating land–sea intermodal strategic corridors into the carbon-trading market remains a critical challenge, particularly due to insufficiently explored game-theoretic mechanisms between governments and liner companies. This study develops an integrated theoretical model combining evolutionary game theory with multiround auctions, uncovering the dynamic strategic interactions between governments and shipping companies in carbon trading. It provides a novel analytical framework and an empirical basis for carbon quota allocation in the New Western Land–Sea Corridor. The main findings are as follows: (1) government intervention is essential for achieving optimal cooperation between governments and liner companies; (2) under government intervention, the optimal number of carbon quota auction rounds is 10; and (3) factors such as carbon quota levels, subsidy amounts, and penalties significantly influence the game outcomes, with carbon quotas being crucial for ensuring the smooth operation of carbon trading. These findings not only address the challenges of integrating carbon-trading mechanisms within the land–sea transport corridor but also offer transferable insights for policy design in similar global corridors (e.g., the Trans-European Transport Network (TEN-T) and the International North–South Transport Corridor), underscoring the necessity of synergistic integration between market mechanisms and government regulation.

Maintaining driving workload (DWL) at an appropriate level is crucial for preventing driver-related crashes. However, the unique conditions of plateau environments significantly impact DWL, increasing driving risks. Research on DWL identification, particularly in real-world plateau driving scenarios, remains limited. This study recruited 27 participants for a naturalistic driving experiment on the Qinghai–Tibet Plateau, integrating psychological and physiological factors to assess DWL. Electrocardiogram (ECG) signals were collected using a wearable wireless physiological monitor, whereas driving video was recorded with two driving recorders. Participants reviewed driving scenarios and operations through recorded videos and rated their subjective DWL using the NASA Task Load Index (NASA-TLX). The self-reported NASA-TLX scores were clustered by C-mean fuzzy (FCM). The cluster results served as classification labels, whereas the corresponding ECG signals were used as features. Then, an extreme gradient boosting (XGBoost) model, optimized by the tree-structured Parzen estimator (TPE) algorithm, classified DWL into three levels. Results show that the proposed model achieves 90.53% accuracy, with an F1 score of 0.91. Under real-world plateau driving conditions, integrating ECG features with subjective workload ratings effectively classified DWL, particularly when using heart rate (HR) and the low-to-high frequency (LF/HF) power ratio. Although the medium level of DWL is more challenging to classify than the other two levels, incorporating multiple physiological features significantly improves the model’s performance in identifying it. These findings provide valuable insights into feature selection and model development for DWL assessment, contributing to optimized road design and enhanced driving safety management in plateau regions.

Blind spots represent a critical challenge to maintaining traffic safety. The development and deployment of intelligent and connected vehicle technologies have resulted in significant enhancements to traffic safety. However, blind spots are still a thorny issue, especially on traffic safety at intersections because of the complexity of their reasoning environment. Despite these advancements, blind spots remain a significant challenge for traffic safety, particularly at intersections, where the complex driving environment hinders accurate perception. In this paper, we introduce an innovative architecture that leverages self-optimizing computational (SOC) network resources to improve the accuracy and efficiency of blind spot detection for vehicles at intersections. This approach tackles two critical challenges: excessive data volumes causing significant transmission delays and insufficient data leading to a deficiency in essential features required for accurate vehicle state assessment. Through dynamic allocation of network resources and real-time performance optimization, it significantly enhances perception and thereby improves traffic safety. (1) Grounded in a comprehensive analysis of real-world conditions, this method enhances performance by focusing on critical areas, optimizing information packaging, and efficiently utilizing communication resources; and (2) this method employs dynamic analysis of blind spots to automatically optimize the allocation of computational resources, ensuring efficient and real-time performance adjustments. To evaluate the proposed perceptual architecture, we validated it using the DAIR-V2X dataset, a benchmark for real-world vehicular infrastructure collaboration, achieving an average precision (AP) of 67.20% at a communication rate of 5.09%.

Currently, the optimization of train stop planning of high-speed railways does not adequately account for how ticket adjustment and train departure time affect passenger expectations. To better align supply and demand dynamics and maximize passenger utility, it is essential to optimize train stop planning, ticket pricing strategies, and passenger flow allocation from a system-wide perspective. The paper proposes a synergistic optimization model for train stop planning, ticket pricing, and passenger flow allocation of high-speed railway. The optimization model aims to maximize the operational revenues of the transportation enterprise while minimizing the total travel costs of passengers. The number of train stops, range of ticket price fluctuations, transportation capacity, and price response function are taken into consideration. A double-layer simulated annealing algorithm is designed to solve the model. Finally, a real case based on the Hohhot East–Beijing North high-speed railway in China verifies the correctness and validity of the model. A comparative analysis of the operational benefits of the railroad transportation enterprises and the travel utility of the passengers under the original plan and the synergistic optimization plan is carried out to verify the validity of the model and the algorithm. The results show that the method proposed in this paper can improve the operational efficiency of transportation enterprises by 18.54% and reduce the passenger travel time costs by 12.13% without increasing the number of trains and the number of stops. The optimized plan can reduce train operation costs, meet passenger flow demand, and improve the operational efficiency of transportation enterprises and the travel utility of passengers.