As automated driving technology advances rapidly, ensuring its safety has become a crucial area of research and practice. This study presents an integrated ensemble learning-logit model (IELLM) designed to enhance the precision in predicting the severity of accidents involving automated vehicles and to explain the factors influencing these accidents. Initially, the model utilizes ensemble learning techniques, integrating multiple machine learning algorithms to predict accident severity and identify key contributing factors accurately. Then, an ordered logit regression model is1 employed to examine critical variables affecting accident severity, such as the level of the automated driving system, pre-collision vehicle speed, and collision site. The findings indicate that vehicles with higher levels of automation (SAE Level 3-5) perform significantly better in accidents than those with lower automation, highlighting the potential of advanced automation technologies to enhance road safety. Moreover, the research identifies high speeds and specific collision types as significant factors contributing to increased accident severity. Through a systematic analysis of these variables, the study deepens the understanding of the characteristics of accidents involving automated vehicles and provides a scientific basis for formulating relevant safety policies and vehicle designs. Based on these findings, recommendations are proposed to improve the safety of automated vehicles, including further research on advanced automated systems, enhanced professional training for drivers, optimized vehicle structural design, and the implementation of stricter speed control measures. These initiatives aim to reduce the incidence and severity of accidents involving automated vehicles, contributing to a safer road traffic environment. (c) 2026 Tongji University and Tongji University Press. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Although human-driven vehicles (HVs) tend to follow autonomous vehicles (AVs) more stably, this stability comes at a cost: HVs must maintain continuous attention and perform frequent micro-adjustments that conflict with their expectations and habitual driving patterns. As a result, driver discomfort accumulates under seemingly smooth traffic conditions, which may paradoxically increase latent risk and contribute to rear-end collisions. We define this phenomenon as non-explicit coordination failure (NECF). Using the Waymo Open Dataset, this study systematically identifies and quantifies the NECF pattern induced by AV driving in no-lane-change, straight car-following scenarios. By constructing a lateral control framework, vehicles traveling in the same lane as the AV (AVLVs) are treated as the affected group, while vehicles one lane away from the AV (SLVs) serve as a weakly affected baseline. We conduct a comparative analysis of longitudinal car-following characteristics and frequency-domain features between these two groups. Furthermore, random forest, logistic regression, and support vector machine models are employed to diagnose whether HVs undergo systematic restructuring under AV-imposed following constraints. The results show that a random forest model achieves approximately 96

Aiming at the limitations of single-dimensional car-following model to describe vehicle car-following behavior in local multi-vehicle environment,the mechanism of vehicles in adjacent lane influencing the subject vehicle car-following behavior is explored,and a vehicle car-following model more suitable for local multi-vehicle environment is attempted to be established.The driving behavior variable reflecting the influence of vehicles in adjacent lane was determined by correlation analysis.Vortisch indicator of similarity(VIS)was used to characterize the influence of vehicles in adjacent lane on car-following behavior.Chi-square independent test and kernel density curve were used to determine the VIS demarcation threshold reflecting whether the influence is significant.Recursive feature elimination was used to screen the variables related to car-following samples significantly affected by vehicles in adjacent lane.The influence mechanism of variables was determined according to the statistical analysis results.Based on the mechanism proposed,the car-following model suitable for local multi-vehicle environment was constructed and its prediction effect was evaluated.Results show that VIS between the speed of subject vehicle and following vehicle in adjacent lane can characterize the influence and the VIS threshold is 0.668.It is concluded that the attention mechanism and memory effect can explainthe influence mechanism of car-following behavior in multi-vehicle environment.Considering the attention mechanism and memoryeffect,RMSE decrease by 65%in full velocity difference model(FVD)model and 62%in intelligent driver model(IDM)model,MAE decrease by 65%and 59%and R2 increased by 180%and 288%respectively,which proved the rationality of the attention mechanism and memory effect in explaining the car following behavior of subject vehicle in local multi-vehicle environment.

The presence of large trucks on highways significantly alters the driving behavior of surrounding vehicles, especially by disrupting the car-following patterns of smaller vehicles due to their size, speed differences, and visibility constraints. This study focuses on investigating the mechanism of how large trucks on highways impact the car-following behavior of surrounding drivers. The research begins by utilizing unmanned aerial vehicles (UAV) to collect vehicle trajectory data. A novel concept, termed the "oppression effects" of large trucks, is introduced, and its influence is characterized using potential field theory. Subsequently, a car-following model is developed that incorporates the oppression effects of large trucks. To illustrate the distribution of these effects, intensity contour maps are employed based on various motion states of the large truck. Finally, the proposed model is then calibrated using real-world trajectory data, and its predictive accuracy is assessed against benchmark car-following models. The proposed model improves trajectory prediction accuracy by over 40.9 % in RMSE and 22.4 % in MAE compared to classical models. The results demonstrate that the car-following model, which accounts for the oppression effects of large trucks, yields more accurate predictions of the driving behavior of vehicles following large trucks on highways. This research contributes to the theoretical foundation for behavior modeling and risk control in mixed traffic environments involving trucks and cars, ultimately enhancing safety for drivers in proximity to large trucks.

Traffic congestion, as a global issue, often leads to adverse social impacts and huge economic losses, especially in urban areas. Utilizing the available urban low-altitude airspace (ULA) is a potential and promising solution to this problem. To fully leveraging ULA and establishing an advanced low-altitude transportation (ALT) system, ensuring the safety of low-altitude flight is of critical importance. However, the ALT system is currently in the exploratory and developmental stage, and the assessment of flight safety relies primarily on pre-flight evaluations and third-party risk indicators. This study introduces a novel flight risk field model considering risk factors during UAV cruising by introducing a new concept of a flight risk field. The model takes into account the key factors influencing the safety of low-altitude flights, considering both the static characteristics of buildings and the dynamic movements of unmanned aerial vehicles (UAVs). It is capable of reflecting the spatiotemporal variations in flight risks during the UAV cruising process. Finally, the model is validated through numerical examples and simulations. The contribution of this paper is to provide a new idea and method for the safety assessment of the ALT system, which can be further applied to airspace structure design, route optimization, and constitution of traffic regulations, to ensure a reasonable airspace design and enhance the safety of low-altitude flight activities.

Automated vehicles are envisioned to navigate safely in complex mixed-traffic scenarios alongside human-driven vehicles. To promise a high degree of safety, accurately predicting the maneuvers of surrounding vehicles and their future positions is a critical task and attracts much attention. However, most existing studies focused on reasoning about positional information based on objective historical trajectories without fully considering the heterogeneity of driving behaviors. Therefore, this study proposes a trajectory prediction framework that combines Mixture Density Networks (MDN) and considers the driving heterogeneity to provide probabilistic and personalized predictions. Specifically, based on a certain length of historical trajectory data, the situation-specific driving preferences of each driver are identified, where key driving behavior feature vectors are extracted to characterize heterogeneity in driving behavior among different drivers. With the inputs of the short-term historical trajectory data and key driving behavior feature vectors, a probabilistic LSTMMD-DBV model combined with LSTM-based encoder-decoder networks and MDN layers is utilized to carry out personalized predictions. Finally, the SHapley Additive exPlanations (SHAP) method is employed to interpret the trained model for predictions. The proposed framework is tested based on a wide-range vehicle trajectory dataset. The results indicate that the proposed model can generate probabilistic future trajectories with remarkably improved predictions compared to existing benchmark models. Moreover, the results confirm that the additional input of driving behavior feature vectors representing the heterogeneity of driving behavior could provide more information and thus contribute to improving the prediction accuracy.

Bridge maintenance is a long-term process that is prone to accidents. Identifying and reducing hidden dangers is crucial in decreasing the occurrence of such accidents. This study proposes a two-stage risk evaluation model based on the likelihood exposure consequence (LEC) method, which includes an occurrence stage and a development stage. The model utilizes hidden danger data accumulated over a long period to reflect the current maintenance stage's risk level. Additionally, a risk prediction model based on the Bayesian network is established to better identify hidden dangers that have a significant impact on construction risk levels (CRLs). The models are validated using 50 weeks of hidden danger data obtained from a real-world bridge maintenance project. The results show that certain hidden dangers have high risk levels when the CRL is high, and small changes in the risk level of certain hidden dangers can have a significant impact on the CRL. This study's models can aid in the development of more targeted HD prevention measures. (c) 2024 Tongji University and Tongji University Press. Publishing Services by Elsevier B.V. on behalf of KeAi Communications Co., Ltd. This is an open access article under the CC BY- NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Hesitant driving behavior is prone to occur in diversion areas where path decisions are required. Significant deceleration or brief stops by hesitant vehicles will reduce the average travel speed and efficiency of the diversion area, increasing the risk of travel in the diversion area and upstream sections. This paper proposes the definition and intensity grading criteria of hesitant driving behavior in the exit diversion area of urban expressway. Firstly, based on the aerial video of UAV and multi-target identification and trajectory tracking algorithm to obtain the vehicle trajectory data of the expressway diversion area, the hesitant driving behavior was divided into two categories based on the characteristics of hesitant behavior and the spatial characteristics of the diversion area, and the hesitant behavior time series feature indicators were proposed; secondly, the hesitant behavior samples were screened using the hesitant behavior feature indicators, and the identification criteria of hesitant driving behavior in the diversion area were established and validated; finally, the hesitation intensity indicators were screened based on systematic clustering, and the intensity grading criteria of hesitant behavior were established by using the isolated forest algorithm combined with the most unfavorable principle. The identification and intensity classification criteria of hesitant driving behavior can provide theoretical basis and technical support for the planning, design and operation management of the expressway diversion area, and guarantee the efficiency and safety of the diversion area.

Spatial analyses of traffic crashes have drawn much interest due to the nature of the spatial dependence and spatial heterogeneity in the crash data. This study makes the best of Geographically Weighted Random Forest (GW-RF) model to explore the local associations between crash frequency and various influencing factors in the US, including road network attributes, socio-economic characteristics, and land use factors collected from multiple data sources. Special emphasis is put on modeling the spatial heterogeneity in the effects of a factor on crash frequency in different geographical areas in a data-driven way. The GW-RF model outperforms global models (e.g. Random Forest) and conventional geographically weighted regression, demonstrating superior predictive accuracy and elucidating spatial variations.The GW-RF model reveals spatial distinctions in the effects of certain factors on crash frequency. For example, the importance of intersection density varies significantly across regions, with high significance in the southern and northeastern areas. Low-grade road density emerges as influential in specific cities. The findings highlight the significance of different factors in influencing crash frequency across zones. Road network factors, particularly intersection density, exhibit high importance universally, while socioeconomic variables demonstrate moderate effects. Interestingly, land use variables show relatively lower importance. The outcomes could help to allocate resources and implement tailored interventions to reduce the likelihood of crashes.

The prediction of operating speed plays a crucial role in road design and safety assessment, especially on complex urban expressway interchange ramps. This task is challenging due to various influences like road conditions, traffic dynamics, and driver behavior. This study aims to identify the optimal model configuration for predicting operating speeds on urban expressway interchange ramps. Three models are established: a short-term operating speed model based on a generalized linear model (GLM), a GLM incorporating for spatial correlation (GLMS), and a deep neural network model considering spatial correlation (DNNS). Each model incorporates considerations for the impact of the plan, profile, and other facets of the interchange ramp in urban expressways. Naturalistic driving experiments are conducted in Shanghai, 70% for model calibration and 30% for validation. Comparative analysis shows that the DNNS model outperforms the others, effectively capturing speed fluctuations along the interchange ramp, demonstrating its robustness and generalization capabilities.

Traditional conflict analysis methods, relying on the assumption of constant velocity, often fall short in capturing the dynamic nature of driver behavior randomness during the interaction process. Predicting all potential collision trajectories proves crucial for comprehensive safety analysis. To address the challenge of accounting for trajectory randomness in car-following scenarios, this study introduces a noise-enhanced generative adversarial network, named Car-Following GAN, designed for predicting collision trajectories based on data from the Shanghai Naturalistic Driving Study (SH-NDS). The model employs an encoder-decoder framework, integrating a noise enhancement module to capture the intrinsic randomness of driving patterns. Demonstrating notable robustness across varying environmental conditions, our model showcases adaptability for trajectory prediction in diverse driving scenarios. A conflict measure, termed the Rear-end Collision Risk Index based on Car-Following GAN (RCRIC), is proposed to quantify the risk of a rear-end collision. Our approach conducts a comprehensive case analysis to assess the impact of various traffic risk factors on RCRIC. The results underscore that our noise-enhanced approach significantly improves the trajectory prediction accuracy of the model when compared to other noise addition methods. This enhancement is observed across various prediction time windows and under different weather conditions. Moreover, RCRIC, derived from the model employing our noise-enhanced approach, effectively mirrors the dynamics of rear-end collision risk by explicitly incorporating trajectory randomness into its assessment. Furthermore, the findings underscore the significant influence of light conditions, traffic density, and weather conditions on driving risk.

Current advanced driver assistance systems (ADASs) do not consider drivers' preferences of evasive behavior types and risk levels under rear-end near-crash scenarios, which undermines driver satisfaction, trust, and use of ADASs. Additionally, spatio-temporal interactions between vehicles are not fully involved in current evasive behavior prediction models, and the influence of evasive behavior is ignored while predicting collision risk. To address these issues, this study aims to propose an ADAS with better driver satisfaction under rear-end near-crash scenarios by establishing a spatio-temporal graph transformer-based prediction framework of evasive behavior and collision risk. A total of 822 evasive events are extracted from 108,000 real vehicle trajectories on highways, and variables from three sources (i.e., road environment features, evading vehicle features, and interactive behavior features) are used to construct rear-end near-crash scenario knowledge graphs (RNSKGs). By utilizing RNSKGs embedding and multi-head self-attention mechanism, spatio-temporal graph transformer networks can effectively capture the spatio-temporal interactions between vehicles. The results show that the prediction accuracy of evasive behavior (i.e., braking-only or braking and steering) and collision risk (lower, medium, or higher risk) is 96.34% and 92.12%, respectively, superior to other commonly-used methods. After including the selected evasive behavior in predicting collision risk, the overall accuracy increases by 10.91%. Then, an autonomous evasive takeover system (AET) based on the prediction framework is developed, and its effectiveness and satisfaction are verified by driving simulation experiments. According to the self-reported data of participants, the safety, comfort, usability, and acceptability of AET proposed in this study all significantly outperform existing autonomous takeover systems (i.e., autonomous emergency braking and autonomous emergency steering). The findings of this study might contribute to the optimization of ADASs, the enhancement of mutual understanding between ADASs and human drivers, and the improvement of active driving safety.

Vehicle trajectory data is in high demand for transportation research due to its rich detail. Lane information is an important aspect of trajectory data, which is typically obtained using sensors such as cameras or LiDAR, which are able to extract road lane features. However, some sensors for trajectory tracking (e.g., MMW radar sensors) are unable to provide lane information. Vehicle detection and trajectory tracking systems based on these sensing technologies can integrate with lane information through manual calibration during initial installation, but this process is labor-intensive and requires frequent recalibration as the sensors gradually become deviated by wind and vibration. This has posed a challenge for trajectory tracking, particularly for real-time applications. To address this challenge, this paper proposes a method for estimating lane-level road geometrics using microscopic trajectory data. The method involves segmenting the trajectory points using direction vectors and clustering them and fitting a series of cluster center points. The mean error (ME) of the distance between the estimated result and the ground truth reference is used to measure the accuracy of the lane-level road geometrics estimation in different conditions. Results show that when the average trajectory data includes at least approximately 30 points per meter in each segment, the ME is always less than 0.1 m. The method has also been tested on MMW wave radar data and found to be effective. This demonstrates the feasibility of our approach for dynamic calibration of road alignment in vehicle trajectory tracking systems.

Accurate driving preferences classification is a crucial component for autonomous connected vehicles in making more safety and more efficient driving decisions. Most existing studies identify drivers’ driving preferences based on the historical data of the individual vehicle, and the selected variables are limited to the mechanical motion of the vehicle, which seldomly takes the influence of road traffic conditions and surrounding vehicles into account. This study proposes a driving preferences classification method by multivariate sequence clustering algorithm based on wide-range trajectory data. Based on the specific range of road sections, the selected variables for each trajectory are converted from the time domain to the space domain separately, to capture the dynamic changes of the features along the road area. Multivariate time series clustering combining a weighted Dynamic Time Warping (WDTW) and the k-medoids algorithm is used to classify driving preferences into different levels, and a popular internal evaluation metric is employed to determine the optimal cluster result. This study also investigates the heterogeneity of driving behaviors at different driving preference levels. The results show that the proposed method could better recognize drivers’ internal driving preferences.

To cope with the randomness derived from the human driving in heterogeneous traffic consists of human-driving vehicles and connected automated vehicles (CAVS), a longitudinal car-following control strategy of CAV is proposed based on the original Intelligent Driver Model (IDM) model and model predictive control (MPC) structure. The string stability of heterogeneous platoon is verified by head-to-tail string stability criteria. Results indicate the strategy proposed can reflect the relationship between the speed and string stability and prove the adaptability to different traffic conditions.

Previous studies on pedestrian crossing have mostly focused on pedestrian crossing decisions; while as an important behavioral aspect, the pedestrian crossing process, i.e., their motions during the entire crossing process, has been narrowly studied. Understanding how pedestrian moves across the street during their entire crossing process helps identify risky movements and reasons for such movements, which can further help in the implementation of effective countermeasures. Therefore, this paper proposed a new and easily applied approach for investigating and understanding the pattern of the pedestrian crossing process at crosswalks based on vision-based trajectory tracking technology and UAV (unmanned aerial vehicle) data. This study uses UAV for collecting video data which is timesaving and has a sufficient coverage area, compared to other methods. For trajectory extraction, the vision-based Deep-SORT-Yolov5 architecture is applied. An improved DBSCAN (density-based spatial clustering of applications with noise) algorithm is introduced for clustering and identification of patterns of pedestrian crossing processes based on their trajectories. This approach is tested via a case study involving six marked crosswalks in Shanghai, China. By using the proposed method, different crossing patterns are extracted and compared. The results show reasonable outputs of trajectory patterns, which reasonably explain the potential instincts of the pedestrians and affecting factors on the behavior of the pedestrian crossing process. Suggestions are made based on the results. This paper contributes to a more comprehensive safety analysis of pedestrian crossings by considering the pedestrian crossing process. The model, along with the UAV-based trajectory observation method, provides an easily-applied and low-cost way of traffic data collection for the purpose of pedestrian safety evaluation.

This paper summarizes the impact of vehicle navigation on driving behavior by searching relevant literature, and looks forward to its new research direction. The current research contents can be summarized as four aspects: navigation mode of vehicle navigation, placement position and screen design, broadcast wording, and prompt timing. The main conclusions are as follows. Among the three navigation methods of visual, auditory and audio-visual combined, the audio-visual combined method has less impact on the overall performance of the driver; the visual and audio-visual combined navigation methods are more effective than auditory navigation,with more glance behavior. When the driver uses vision for navigation, the driver’s gaze time on the front and left side of the road will decrease; the screen design will affect the driver’s interaction time with the navigation screen. Setting more detailed and specific broadcast words such as distance information, direction information, road information, and lane information is more conducive to drivers’ understanding and can improve driving stability. The prompt timing based on speed setting is more superior, the driver’s reaction is faster,and the vehicle stability is better.

为探究驾驶人速度感知机理,设计驾驶模拟实验采集驾驶人感知速度与实际行车速度,分析地下快速路几何线形、侧壁变化频率等因素对驾驶人速度感知敏感性的影响,并基于恒定速度理论、机器学习方法构建驾驶人速度感知偏差敏感性阈值预测模型.结果表明,在0.05显著水平下,不同速度水平、线形组合、侧壁变化频率组别下,驾驶人的速度感知敏感性阈值都存在显著差异.采用LASSO(least absolute shrinkage and selection operator)筛选根据短期、中长期、长期时窗划分的各特征变量,构建的多元非线性模型的决定系数R2为0.645,多层感知机模型的决定系数R2为0.727,支持向量回归模型的决定系数R2为0.853.

With the increase in the scale of urban underground space development, the issue of road safety has received more attention. Compared with highways and urban surface roads, underground roads lack a mature driving safety evaluation index system, and the current driving safety evaluation does not consider the impact of traffic flow. In the context of smart transportation, this study focuses on the driving safety level of urban underground roads. Based on the existing research results of driving safety evaluation indicators,the data granularity is refined from the road section unit to the index data change of the time window of 1s to mine underground traffic.Based on the vehicle running characteristics and traffic conflict characteristics of the continuous section of the road, an underground road driving safety evaluation index system is constructed considering the traffic flow state, and the threshold value of the evaluation index suitable for the underground road environment is determined by the exceeding threshold method in the extreme value theory.

针对交通强国建设对道路工程专业人才培养的需求,在现有传统理论教学基础上,探索开展了基于模拟驾驶的超车换道意图预测实验教学项目.实验对象为高速公路典型场景下驾驶行为特性,实验课程包含理论分析、场景构建、模拟驾驶、实验数据采集、数据分析、预测建模、讨论交流等环节,通过该实验教学项目,能够加深学生对道路安全、驾驶行为理论知识的理解与应用,锻炼使用先进科学仪器研究问题和解决问题的能力,培养学生的科学研究兴趣,进一步加强道路工程实践创新人才的培养.