交通信息与安全

2025, 43(2): .

Abstract(28) HTML (22) PDF(2)

Abstract:

An Analysis of the Seafarers' Unsafe Actions Causality Network Based on Association Rule Mining

MA Xiaoxue, ZHANG Ruiwen, QIAO Weiliang, HAN Bing, YANG Jie

2025, 43(2): 1-10. doi: 10.3963/j.jssn.1674-4861.2025.02.001

Abstract(41) HTML (14) PDF(13)

Abstract:
Seafarers' unsafe actions are pivotal contributors to the frequent occurrence of waterborne traffic accidents. However, the majority of existing research tends to concentrate on the analysis of single factors, with insufficient exploration of the underlying mechanisms involving the interplay of multiple factors. Based on 886 waterborne traffic accident investigation reports, unsafe actions and its causal factors are extracted by utilizing Grounded Theory. According to the framework of human factor analysis and classification system (HFACS), an analytical framework for the unsafe actions and causal factors is established, encompassing five levels and 76 factors. association rule mining (ARM) algorithm is utilized for the exploration of the coupling relationship and interaction between the seafarers' unsafe actions. It reveals how various causal factors synergistically contribute to the occurrence of seafarers' unsafe actions. By employing complex network theory, the results of association rule analysis are mapped onto a directed weighted network, constructing a causal network model for seafarers' unsafe actions. The key nodes influencing seafarers' unsafe actions are identified by analyzing the topological characteristics of the network. It is highlighted that the causation network of seafarers' unsafe actions exhibits typical small-world network characteristics, with an average clustering coefficient of 0.63 and an average path length of 2.095 2. This indicates that the influencing factors are closely interconnected, making it susceptible to triggering chain reactions. Among the strong association rules, "severe lookout negligence by watchkeeping seafarers" has a 55% probability of interacting with other factors to trigger other seafarers' unsafe actions. The betweenness centrality value of it is approximately 0.262 7, playing a crucial intermediary role in the development of causation pathways. "Failure to adopt a safe speed" and "inadequate use of navigational aids" exhibit a highly correlated relationship in grounding accidents, with a mutual inducement probability of 70% when interacting with other factors. Unsafe actions such as "failure to detect and implement effective risk mitigation measures in a timely manner" and "poor emergency response capabilities of the captain" frequently emerge in the association rules across multiple accident types. It indicates that these unsafe actions occupy a pivotal position in the process of seafarers' risk management.

Operational Risk Assessment of Terminal Airspace with Prevailing Traffic Flow

ZHI Jingwen, ZHANG Junfeng, MA Zao

2025, 43(2): 11-18. doi: 10.3963/j.jssn.1674-4861.2025.02.002

Abstract(37) HTML (19) PDF(0)

Abstract:
The characterization and assessment of operational risks represent critical steps in enhancing airspace operational safety and management capabilities, particularly in structurally complex terminal airspaces with multiple intersecting arrival and departure flows. Traditional collision risk models predominantly focus on preemptive single-aircraft conflict detection and fixed-routes, exhibiting limitations in characterizing terminal airspace operational structures and analyzing post-conflict risks. This study develops a methodology that extracts prevalent traffic flows from historical trajectory data through clustering, which identifies geometric configurations of conflict scenarios and incorporating aircraft maneuvering adjustments. By analyzing the spatiotemporal distribution in risk-prone areas, an assessment metric defined as the expected collision frequency per unit time was constructed to quantify the operational risk. Cumulative aggregation across diverse temporal and spatial scales enables risk quantification and hotspot identification, thereby facilitating operational evaluations, early warnings for high-risk traffic flows and spatiotemporal hotspots, and providing guidance for air traffic control, tactical planning, and airspace construction. A case study of Guangzhou Baiyun Airport validates the model's effectiveness in risk characterization and hotspot detection, Key findings include: Conflict risks between arrival flows constitute the primary risk category; The departure flow of VIBOS requires attention under northbound operation; Temporal risk hotspots are concentrated between 13:00—14:00 and 22:00—02:00.

An Empirical Study on the Impact of Information Content of Traffic Signs in Entrance Areas of Highway Tunnels on Drivers' Visual Behavior

JIANG Luqing, DU Zhigang, MAI Jing

2025, 43(2): 19-27. doi: 10.3963/j.jssn.1674-4861.2025.02.003

Abstract(31) HTML (14)

Abstract:
To evaluate the impact of information content of traffic signs in entrance areas of highway tunnels on drivers' visual behavior, this study conducted a real-driving experiment, combined with eye movement entropy and a coupling coordination degree model, with which the influence patterns of different levels of information content on drivers' visual behaviors is explored and the coordination mechanism between information content and characteristics of eye movement is revealed. The experiment selected six entrance areas of highway tunnels, representing six levels of information content of traffic signs (i.e., T0—T5, ranging from 0 to 81.18 bits). Forty drivers were recruited wearing eye trackers to collect eye movement data. Metrics such as fixation duration, saccade duration, and saccade amplitude were analyzed. Eye movement entropy was calculated based on sample entropy theory to analyze the complexity of visual search patterns, and a coupling coordination degree model was constructed to assess the coordination level between information content and eye movement behavior. The results indicated that the information content of traffic signs in the entrance areas of highway tunnels significantly affected drivers' eye movement characteristics. As the information content increased, fixation and saccade durations first decreased and then increased, while saccade amplitude first increased and then decreased. The T3 level (i.e., 48.31 bits) showed the best performance across all metrics, reflecting the highest efficiency in drivers' information perception and search. As vehicles approached the tunnel, the sample entropies of fixation duration, saccade duration, and saccade amplitude all gradually increased, and the rate of growth within the recognition sight distance range (125—100 m) significantly increased, indicating that drivers' search intensity for environmental information increased as the distance shortened. At the T3 level, eye movement entropy was the smallest, and the visual search pattern was the most stable. The coupling coordination degree between information content and eye movement behavior exhibited a unimodal curve that first increased and then decreased. The coupling coordination degree at the T3 level reached 0.851, which falls in a "good coordination" level (Level 9), while both levels of insufficient information (T0—T1) and information overload (T4—T5) resulted in a state of imbalance.

A Method of Risk Assessment for Subway Stations Based on D-S Evidence Theory

DONG Sheng, MA Yunjie, ZHOU Jibiao, DU Yunchao, LI Zewei

2025, 43(2): 28-35. doi: 10.3963/j.jssn.1674-4861.2025.02.004

Abstract(33) HTML (20)

Abstract:
The safety risk assessment of high-density passenger flow holds significant importance for improving the emergency response capabilities of urban metro systems. To address t traditional methods'limitations, such as inadequate indicator systems, insufficient integration of multi-source data, and low assessment accuracy, an enhanced Dempster-Shafer (D-S) evidence theory method was proposed. This method is structured around a five-dimensional indicator system encompassing personnel, equipment, passenger flow, environmental conditions, and management protocols. Comprehensive weights were determined through Game Theory-Combinatorial Empowerment. The safety level membership degrees of each indicator were quantified using a fuzzy half-gradient affiliation function, while an evidence similarity matrix is constructed via Jousselme's evidential distance function. To address high-conflict evidence, a calibration factor α, and adjustment parameter μ, are introduced to refine the fusion process, where the final risk level derived through a linear weighted method. A case study was conducted at Ningbo Metro Gulou Station, utilizing holiday evening peak passenger flow data and expert evaluations to establish a multi-source evidence set for validation. The results demonstrated that: ①Compared to the traditional D-S method and Yager's method, the proposed approach reduces average evidence conflicts by 34.4% and 8.5%, respectively; ②The membership grade of the critical"passenger flow"indicator has an R3 rank value reaches 0.8202, confirming the method's effectiveness in characterizing scenarios of high-density passenger flow; ③The proposed method exhibits robust adaptability and stability, with an error rate below 5% in the comparison of multiple scenarios. These findings provided actionable insights for identifying and mitigating risks in metro systems under high-density passenger flow conditions.

A Method for Inland Vessel Object Detection Based on PEW-YOLOv8

CAO Zhiyuan, MA Yong, CHENG Xuefu, HU Wentao

2025, 43(2): 36-43. doi: 10.3963/j.jssn.1674-4861.2025.02.005

Abstract(37) HTML (23)

Abstract:
In inland vessel object detection, many targets fall into the category of small objects, occupying limited pixels in images. Additionally, interference from complex environments often leads to insufficient detection accuracy, frequent false positives, and missed detections. To address these challenges, this study proposes an object detection algorithm based on PEW-YOLOv8, which integrates YOLOv8 with a P2 detection layer, EfficientNetV2, and the WIoUinner loss function. A new P2 shallow detection layer with a resolution of 160×160 is introduced to enhance small target detection. A 32-dimensional feature space reconstruction is employed to achieve dynamic weight allocation across multi-scale features. Furthermore, a bidirectional interaction mechanism between high- and low-level features is designed to improve feature extraction for small vessel objects. To address the increased parameter burden caused by multi-level detection heads, an improved EfficientNetV2 architecture is adopted, which incor-porates a GELU-activated Stem module to mitigate gradient explosion and unstable training. During training, the channel count is expanded fourfold while simplifying the convolutional structure, significantly accelerating the training process without sacrificing model quality. Besides, the WIoUinner loss function with a dynamic non-monotonic focusing mechanism is designed, which introduces auxiliary prediction boxes with varying scales to accelerate the convergence of bounding boxes. When the predicted and ground truth boxes are closed aligned, the model places greater emphasis on the distance between center points, reducing the penalty from geometric metrics and improving generalization capability. The algorithm is validated using a dataset that combines the publicly available Seaships dataset with a self-constructed inland vessel dataset. Experimental results demonstrate that compared to YOLOv10, PEW-YOLOv8 achieves an average detection accuracy of 94.8%, a 3% improvement. Computational complexity is significantly reduced, with FLOPs optimized to 3.7 G, representing a 43.1% reduction, which demonstrates the model's advantages in both accuracy and efficiency for inland vessel detection tasks. Heatmap analysis further confirms the model's ability to effectively focus on inland vessel features, demonstrating robust detection performance in complex inland waterway scenarios.

A Model for Aircraft Surface Taxiing Trajectory Prediction Base on Transformer-TCN-GRU

WANG Xinglong, LI Guoxiang, ZHANG Zhao, YE Ke, SU Ting, GE Jing

2025, 43(2): 44-53. doi: 10.3963/j.jssn.1674-4861.2025.02.006

Abstract(58) HTML (26)

Abstract:
For aircraft taxiing trajectory prediction, existing methods exhibit low accuracy in real-time estimation of future positions over medium-term time horizons. To enhance prediction precision within this temporal scope while maintaining computational efficiency, this study proposes a taxiing trajectory prediction model integrating transformer networks, cross-attention mechanisms, temporal convolutional networks (TCN), and gated recurrent units (GRU) to generate multiple candidate trajectories. The Transformer encoder captures temporal dependencies and motion patterns from historical trajectory data to derive global feature representations. Airport vector maps and taxiing route instructions from air traffic control systems are utilized to compute planned future taxiing path coordinates. A cross-attention mechanism then aligns the global trajectory features (as Query) with critical positions in the planned path sequence, mapping the fused path-enhanced features into multimodal representations corresponding to candidate trajectories. The TCN-GRU decoder processes each modality to capture long-term temporal dependencies and outputs multiple predicted trajectories with associated probabilities. Validation on real taxiing trajectories from a major Chinese airport demonstrates minimum average displacement error (minADE) of 1.932 m and minimum final displacement error (minFDE) of 1.811 m for 8-second predictions. Compared to individual GRU and TCN models, the proposed approach reduces minADE/minFDE by 14.10%/30.88% and 16.62%/34.72% respectively, while maintain an average runtime of 17.70 milliseconds per sample. The proposed method achieves accurate and efficient trajectory prediction, supporting enhanced safety management in airport maneuvering areas.

A Detection Method for Road Surface Pothole Based on Mobile-scanned Point Cloud Using Graph Neural Networks

ZHANG Tingrui, ZHANG Xuequan, YANG Zichuan, MA Wenshuo, LIU Bing

2025, 43(2): 54-64. doi: 10.3963/j.jssn.1674-4861.2025.02.007

Abstract(43) HTML (29)

Abstract:
Rapid detection and assessment of road surface potholes are essential for traffic safety. To address the high cost and limited applicability of current detection methods based on survey vehicles or drones, as well as the low quantitative accuracy of smartphone-based approaches, this study proposes a novel method for pothole extraction and quantification from smartphone-scanned point clouds using a graph-based attention neural network (GANN). Road surface point cloud data are collected using a LiDAR-equipped smartphone via circular scanning and are preprocessed through planar fitting and clustering algorithms. To effectively capture the local geometric features characteristic of pothole structures, a deep learning model is developed based on graph attention mechanisms, extending traditional graph neural network (GNN) models. The proposed GANN model introduces an Attention Neighbor Convolution Layer, which identifies key neighboring nodes within an expanded receptive field using attention mechanisms, addressing limitations associated with dynamic graph construction present in existing approaches. Additionally, a Geometric Feature Extractor is designed by incorporating an umbrella surface representation to accurately characterize local geometric structures that are often overlooked by prior methodologies. These architectural enhancements enable high-precision classification and quantitative analysis of the preprocessed point cloud data. Experiments were conducted using an iPhone 14 Pro to scan road surface potholes around the Yujiatou Campus of Wuhan University of Technology in Wuchang District, Wuhan, resulting in a real-world urban road pothole point cloud dataset. Results show that the proposed GANN model achieves a depth quantification error of 4.58% and a volume quantification error of 5.57%, demonstrating its effectiveness in extracting potholes from point cloud data. Compared with state-of-the-art models such as PointNeXt and PointMLP, GANN reduces depth and volume quantification errors by 2.41% and 0.11%, respectively, offering superior accuracy in pothole quantification through improved information retention and geometric feature extraction.

A Model for Predicting Ship Emission Pollutants Based on MASTGCN Using AIS Information

YAO Danyang, YUE Mingqi, ZHANG Xun, WU Fang, CHENG Shiming

2025, 43(2): 65-73. doi: 10.3963/j.jssn.1674-4861.2025.02.008

Abstract(26) HTML (16)

Abstract:
Sulfur dioxide (SO₂) emissions from ships are a major contributor to air pollution and ocean acidification, exhibiting significant spatial and temporal heterogeneity. Current prediction models for shipborne pollutants have limitations in modeling spatiotemporal dependencies, making it difficult to effectively capture the complex spatiotemporal correlation characteristics in SO₂ emissions. To address this issue, based on automatic identification system (AIS) data and Chinese ship registry data, a dynamics-based method combined with emission factor approaches is used to quantify shipborne SO₂ emissions during navigation, thereby providing a solid data foundation for subsequent prediction. In terms of model construction, a multi-head attention spatial-temporal graph convolutional network (MASTGCN) is proposed. Based on the spatial-temporal graph convolutional network (STGCN) architecture, MASTGCN incorporates multi-head self-attention mechanisms in both spatial and temporal dimensions. By dynamically allocating weights, it enhances the modeling capability to learn spatial dependencies across different regions and temporal dependencies across time intervals, thus improving the accuracy of spatiotemporal predictions for shipborne SO₂ emissions. Experimental results show that when the number of attention heads is set to five, the model achieves a mean absolute error (MAE) of 0.057 5, mean squared error (MSE) of 0.120 6, root mean squared error (RMSE) of 0.347 3, and floating point operations (FLOPs) of 3 030 M. These results demonstrate superior overall performance in both accuracy and efficiency compared to other configurations and the baseline STGCN model. Specifically, MASTGCN with five attention heads outperforms STGCN by improving MAE by 27.6%, MSE by 6.0%, and RMSE by 1.3%. The findings indicate that the incorporation of multi-head attention mechanisms enables the model to effectively capture the spatial characteristics of SO₂ emissions through dynamic weighting. The five-head MASTGCN model achieves excellent predictive accuracy while maintaining a relatively reasonable computational complexity.

A Method for Dynamic Coupling Coordination of High-speed Railway Node-line Passing Capacity Based on SEM

ZHANG Chunmin, ZHU Yuanjing, JIANG Yuxing

2025, 43(2): 74-84. doi: 10.3963/j.jssn.1674-4861.2025.02.009

Abstract(28) HTML (14)

Abstract:
This study addresses the challenges posed by multiple dynamic factors affecting the passing capacity of high-speed railway node-line systems, particularly focusing on the complex interrelationships among these factors and the difficulty in quantifying their weights. A dynamic coupling coordination method based on structural equation modeling (SEM) is proposed to resolve these issues. The dynamic factors influencing high-speed railway node-line passing capacity are categorized into six subsystems: node system, line system, transportation organization, operational time, delays, and early arrivals. A PLE-SEM-based dynamic coordination model is subsequently constructed to analyze these subsystems. Operational data from the Shanghai West Station-Shanghai Station section are utilized to verify the model's validity, scientific rigor, and goodness-of-fit through measurement model and structural model validation. This process identifies interaction mechanisms, directional influences, and indicator weights across subsystems while detecting capacity bottlenecks. A coupling coordination degree model is further applied to assess system-level and subsystem-level coordination states, enabling the identification of key links. The results demonstrate: ①this model method is applicable for investigating the complex influence relationships among dynamic factors affecting the throughput capacity of high-speed railway node-line systems. It enables the identification of critical dynamic factors within subsystems of the system's throughput capacity, quantifies the influence relationships, and specifies the direction of impacts. ②Through comprehensive factor analysis and case study calculations, the proposed method effectively identifies key factors and reveals dynamic coupling relationships among them. A total of 13 direct influence paths and 9 indirect influence paths are established between subsystems, visualizing their interconnections. These results provide a basis for selecting appropriate capacity calculation parameters in practical applications.

Urban Adaptive Travel Hotspot Detection and Area Division Method

YAN Yuchen, WANG Yuxin, QUAN Wei

2025, 43(2): 85-94. doi: 10.3963/j.jssn.1674-4861.2025.02.010

Abstract(38) HTML (22)

Abstract:
To address the limitations of fixed bandwidth hotspot detection in multi-density travel data and the spatial heterogeneity distortion from traditional zoning methods, this study proposes an adaptive hotspot detection and dy-namic regional division method for urban transportation. An adaptive travel density estimation model is developed. A global initial bandwidth is determined via pilot estimation, and sensitivity parameters are calibrated using maxi-mum likelihood estimation. Local bandwidth correction factors and a dynamic bandwidth adjustment mechanism en-able automatic research bandwidth regulation. A multilevel hotspot identification technology is developed, combin-ing moving window extreme value detection with natural breaks classification to form a travel hotspot evaluation system. Furthermore, Voronoi polygons are generated using the hotspots as control points to serve as basic analysis units while preserving spatial heterogeneity characteristics. The effectiveness of regional division is evaluated using 5 indicators, including travel heat, Moran's index, and others. Empirical analysis using Harbin's main urban area taxi trajectory data shows that, compared with fixed bandwidth methods, the proposed method increases identified hotspots by 2.1 to 6.7 times. The standard deviation of regional travel heat differences is 572.8. The average dis-tance between point data centroids within areal elements and the geometric center is 137.8 m, a 15.1% to 74.3% re-duction from traditional grid methods, verifying the homogeneity advantage of the regional division. The nugget to sill ratio drops to 0.135, and internal unit variation decreases by 39.6%, indicating the method effectively retains da-ta aggregation characteristics and reduces the modifiable areal unit problem's impact. The adaptive method identifies 1, 719 travel hotspots in the study area and accurately locates road intersection hotspots in areas like Harbin West Station, with clear boundaries and explicit geographical semantics. The results provide an adaptive framework for multi-density travel data analysis, supporting applications such as taxi dispatching and demand forecasting.

A Merging Model Based on Piecewise Deep Reinforcement Learning for Connected and Autonomous Vehicle in Work Zone under Mixed Autonomy

XIN Qi, JIA Shengqi, XU Meng, QI Jiale, YUAN Wei

2025, 43(2): 95-108. doi: 10.3963/j.jssn.1674-4861.2025.02.011

Abstract(34) HTML (29)

Abstract:
The classical early and late merge work worse under dynamic demand, and render conflict merging gap due to large speed differences at the upstream. To this end, a piece-wise deep reinforcement learning-based merging model is proposed for connected and autonomous vehicles (CAVs) in work zones under mixed autonomy. Above all, the merging conflicts and efficiency reduction caused by many vehicles in closed lanes trying to merge into one gap on the open lane are addressed by the model with speed guidance, gap creation, and positional alignment. Such a model consists of the soft Actor-Critic algorithm-based longitudinal control and the rule-based lane-changing decision-making. For longitudinal control, 9 features are selected as the agent state to describe surrounding traffic conditions from both local and global views. The mentioned features include the speed and acceleration of the ego vehicle, the speed of and the distance to the lead vehicle, the speed of and the distance to the lead and lag vehicles on the adjacent left lane, and the distance to the merging point. Subsequently, a piecewise reward function for CAVs in the work zone is established by optimizing comfort, safety, and efficiency simultaneously. Such a reward function combines minimizing acceleration and jerk, preventing collisions, generating merging gaps, aligning with the gap center on the open lane, mitigating vehicular speed differences, adhering to advisory speed, and encouraging following vehicles with yield behavior. Particularly, an item of reward function with respect to driving efficiency is shaped on the basis of the speed difference between the lag vehicle on the adjacent lane and the ego vehicle, such that halting of both the CAV and the human-driving vehicle can be alleviated at the merging point. Simulation results illustrate that the proposed model increases by about 4.76% of average speed, and 19.71% of minimal time-to-collision under medium/heavy demand in work zone, in contrast to early merge, late merge and New England merge. In addition, the average speed, minimum time-to-collision, and successful merging rate in mixed autonomy with heterogeneous human-driving vehicles, increase with the increase of the CAV market penetration rate, while all the vehicles merge without halting.

A Study on Optimization of Operation Procedures in the Cul-de-sac Area of Terminals in Large Hub Airports

OUYANG Jie, SUN Mingyang, ZHU Changqing, KOU Weibin

2025, 43(2): 109-118. doi: 10.3963/j.jssn.1674-4861.2025.02.012

Abstract(28) HTML (14) PDF(0)

Abstract:
Large hub airports have widely adopted the finger corridor terminal configuration to obtain more contact stands. However, due to limited spatiotemporal resources and an imbalance in supply and demand, the U-shaped Cul-de-sac area formed between finger corridors has gradually become a bottleneck for efficient airport operations. To address restricted entry and exit of aircraft in the Cul-de-sac areas during peak hours, this study proposes optimization procedures for aircraft operations. By analyzing the structural characteristics and operational status of the Cul-de-sac areas, transportation engineering technologies are developed, and the formation operation procedures are proposed based on the current status and engineering-optimized technology. Considering the operational rules of aircraft in the Cul-de-sac area, an optimization model for the entire operational process is established, and a dynamic marshalling algorithm incorporating safety distance factors is designed to enhance efficiency. To compare operational efficiency before and after the optimization, validation is conducted using actual flight data from the Cul-de-sac area on the northwest side of Guangzhou Baiyun Airport. Three operational scenarios are compared: actual operations, formation operations and engineering-optimized formation operations. The results show that, compared with the baseline operations, the average operation time of full-time flights is reduced by 11.37% under formation operations and by 14.45% under engineering-optimized formation operations. Specifically, the average operation time of departure flights is reduced by 6.47% and 10.13%, while that of arrival flights is reduced by 20.27% and 22.31%. Moreover, flight delays are reduced by 45.94% and 58.42% respectively. Based on the current operations and the op-timized engineering technology, the number of flight groupings during the entire period is 29 and 34, respectively. The maximum number of flight groupings during peak hours is 3, which verifies the effectiveness of optimizing the flight grouping procedure in improving overall operational efficiency in the Cul-de-sac area under complex configu-rations and scenarios.

An Analysis of Factors Influencing the Willingness to Use Automated Driving Function

SUN Shouzhong, QIN Hua, CHEN Qixuan, RAN Linghua

2025, 43(2): 119-126. doi: 10.3963/j.jssn.1674-4861.2025.02.013

Abstract(23) HTML (17) PDF(0)

Abstract:
With the accelerated launch of automated driving function (ADF) vehicles on the market, the actual usage rate of its users has shown a low trend. In order to promote the acceptance and application of ADF technology by drivers, it is crucial to analyze the key factors affecting their willingness to use it. Previous studies have examined drivers' willingness to use automation functions in vehicle. However, due to technological limitations at the time, surveyed individuals generally lacked adequate practical experience. In view of this, this study conducts an extensive questionnaire survey for fully experienced user groups. This paper explores the key factors affecting the will-ingness to use from three types of user information: demographics, behavior patterns and feeling evaluation. Based on the literature and established scale, the study devises a questionnaire concerning the willingness to use of Automated Driving Function (ADF). It collects 223 valid questionnaires via online and offline methods. The prediction model of ADF willingness to use is constructed through correlation analysis and hierarchical regression analysis, and the influence of three types of user factors on users' willingness to use is explored. The results show that : ①in the current environment, the constructed predictive model for willingness to use ADF can explain 68.9% of the variance. ②Perceived safety stands out as the predominant forecasting predictor, accounting for 36.2% of the variance in willingness to use. ③New technology orientation, perceived usefulness, trust, understanding and age also have a significant impact on the willingness to use, among which new technology orientation is the biggest factor affecting the willingness to use in behavioral pattern information.④Although the user's behavior pattern has a significant impact on the willingness to use autonomous driving function, it can still improve the willingness to use through a benign driving experience.

Predicting the Demand of Ride-Hailing Based on a QRF-SA-ConvLSTM Model

LI Xiaomei, LYU Bin

2025, 43(2): 127-135. doi: 10.3963/j.jssn.1674-4861.2025.02.014

Abstract(24) HTML (16) PDF(0)

Abstract:
Considering the spatiotemporal features and predictive uncertainty in the demand prediction of ride-hailing are essential for improving the efficiency and robustness of the transportation system. This study proposes a novel prediction framework that integrates quantile regression forests (QRF) and a self-attention convolutional long short-term memory network (SA-ConvLSTM) to jointly optimize spatiotemporal feature learning and probabilistic modeling, which could lead to high-precision prediction and uncertainty estimation. First, the model employs QRF to generate the distributions of predicted demand within specified probability intervals from historical spatiotemporal data. It then establishes a spatiotemporal probability matrix by weighted fusion of original data through the attention mechanism. The matrix is then fed into the SA-ConvLSTM module, where the convolutional structure extracts local spatial patterns, the self-attention mechanism focuses on key spatiotemporal nodes, and LSTM captures long-term temporal dependencies. Finally, a multi-task output layer simultaneously optimizes point predictions and interval estimations. Combining the ability of distributional modeling of QRF and the advantage of extraction of spatiotemporal feature of deep learning, the proposed model improves both prediction accuracy and uncertainty estimation. The study validates the model's effectiveness and accuracy using Didi's open-source dataset of ride-hailing trajectories within the second ring road in Xi'an. The results show that in terms of mean prediction, the proposed model reduces mean absolute error and mean squared error by 21% and 17.2%, respectively, comparing to the SA-ConvLSTM baseline; in terms of uncertainty estimation, the prediction interval coverage probability at 95% confident level improves by 5.3%, 2.5%, and 0.9% comparing to linear quantile regression (LQR), QRF, and SA-ConvLSTM, respectively, while the average width of the prediction results is reduced by 41.5%, 30%, and 18.5%. These results validate the proposed model's goodness of fit in terms of both predictive accuracy and reliability.

2025, 43(2): 136-136.

Abstract(17) HTML (12) PDF(0)

Abstract:

A Review of Traffic Operation Risk Analysis on Highways in a Connected and Automated Environment

ZHANG Mengya, YANG Xiaoguang, MA Chengyuan, YANG Jie

2025, 43(2): 137-153. doi: 10.3963/j.jssn.1674-4861.2025.02.015

Abstract(43) HTML (32) PDF(0)

Abstract:
The analysis of traffic operation risks on expressways holds significant theoretical and practical value for enhancing traffic safety, reliability, and realizing proactive management. The connected environment introduces new methods, theories, and technologies for traffic risk analysis. Focusing on continuous traffic flows on highways and urban expressways, this study systematically reviews the theoretical foundations and key techniques of traffic operation risk analysis under both non-connected and connected environments, and discusses future research directions. In non-connected environments, risk analysis primarily relies on limited unstructured data, emphasizing risk factor identification, accident mechanism analysis, and risk prediction. Challenges remain in achieving high-precision scene modeling, responsive dynamic evolution analysis, and real-time risk perception and forecasting. In connected environments, the integration of multi-source real-time data enables a shift toward proactive risk prediction and localized interaction analysis, with improvements in data support and modeling accuracy. However, in mixed traffic conditions, systematic models for the interaction dynamics of heterogeneous vehicle types have not yet been fully established. The combined effects of diverse driving behaviors, communication delays, and perception deviations require further investigation, and the modeling and interpretation of dynamic factor evolution in complex environments remain incomplete. Future research should advance the modeling of human-vehicle-road collaborative evolution, refine the characterization of risk accumulation and propagation in mixed traffic, enhance multi-source data fusion and dynamic feature extraction, and strengthen the real-time performance and robustness of risk evaluation. Through theoretical innovation, data-driven methods, and integrated technological development, it is expected that an interpretable, predictable, and actionable intelligent traffic risk prevention and control system will be progressively established.

A Review of Pavement Distress Detection Based on Machine Learning Methods

ZOU Zheng, CHEN Jiang, LANG Hong, WANG Xiaofeng, WAN Chenguang, DING Shuo, LU Jian

2025, 43(2): 154-168. doi: 10.3963/j.jssn.1674-4861.2025.02.016

Abstract(43) HTML (20) PDF(0)

Abstract:
Road surface quality directly influences service life and driving safety. Pavement distress involves complex mechanisms, diverse forms, and large-scale spatial distribution, posing challenges for measurement and annotation. Traditional detection methods struggle to meet practical needs for identifying various distress types in diverse scenarios. This paper reviews achievements in machine learning-based pavement distress detection, compares principles and applicability of recognition methods, and summarizes public and private image datasets of pavement distress. Subsequent feature analyses of pavement distress using 3D depth data provide a basis for exploring interactions between multidimensional features and AI models. Finally, general objective evaluation metrics are systematically introduced to ensure fairness in AI-based pavement distress assessment. Distress recognition has evolved from threshold segmentation, element classification, and object detection to pixel-level segmentation, significantly improving accuracy and generalizability. Traditional machine learning and deep learning methods can be integrated to detect more distress types and learn more effective features, improving accuracy and efficiency. Current algorithms should be evaluated using consistent metrics, including runtime, memory usage, computational load, and recognition rate, while considering datasets with varying distress dimensions and road environments. Future work should develop detection and processing methods for complex conditions, enhance signal-to-noise ratios and generalization, and support deploying intelligent algorithms in diverse scenarios.

A Roadside Multi-sensor Position Calibration Method for Wide-area Perception

LI Chengmin, WANG Junhua, FU Ting, SHANGGUAN Qiangqiang

2025, 43(2): 169-176. doi: 10.3963/j.jssn.1674-4861.2025.02.017

Abstract(44) HTML (18) PDF(0)

Abstract:
Accurate sensor calibration within the stake number coordinate system is essential for wide-area perception in the development of intelligent highway infrastructure. However, traditional calibration methods still struggleto meet the required accuracy and involve low efficiency and safety concerns due to manual stake number recording. To address these limitations, this study proposes a calibration method of roadside multi-sensor position using atest vehicle equipped with real-time kinematic (RTK) devices. The vehicle conducts fixed-point measurements ateach sensor location to obtain high-precision geographic coordinates (latitude and longitude), which are subsequently aligned with manually recorded stake numbers to facilitate coordinate calibration. An engineering coordinate system is introduced as an intermediate reference, enabling a unified transformation from latitude and longitude coordinates to engineering coordinates and ultimately to stake number coordinates. To improve robustness and automation, the proposed method incorporates an enhanced random sample consensus (RANSAC) based algorithm, which leverages prior knowledge of road geometry and incorporates a parameter adaptation mechanism. Calibration accuracy isfurther optimized through an automatic threshold selection strategy guided by mean error metrics and inlier variation rates, allowing for the detection of abnormal stake numbers and the exclusion of outliers. Experimental resultsshow that the proposed method achieves a calibration error of 0.28 m, successfully identifying and correcting 5 outliers, significantly outperforming the traditional least squares method, which yields a 0.63 m error and fails to identifyany outliers. In comparison, the truncated least squares method results in a 0.35 m error with 21 inliers, while theleast median of squares method achieves a lower error of 0.19 m but retains only 14 inliers. The proposed methodmaintains 19 inliers, balancing calibration accuracy and data retention, and achieves a superior trade-off between accuracy and robustness. Based on the calibrated sensor positions, wide-area trajectory data aligned in the stake number coordinate system can intuitively represent lane-level vehicle dynamics and stake number information, demonstrating the method's practical applicability and scalability in intelligent transportation systems.

A Method for Traffic Risk Identification Under Complex Weather Conditions Based on the BKPE Security Field Model

LI Chengxin, LIU Benmin, LIAO Chenfei, WANG Pengfei, HU Jiaxin, LIU Pengqian, TU Huizhao

2025, 43(2): 177-186. doi: 10.3963/j.jssn.1674-4861.2025.02.018

Abstract(36) HTML (20) PDF(3)

Abstract:
The current driving safety field (DSF) theory is based on a three-dimensional framework of "driver-vehicle-road" to construct the potential energy function. However, it overlooks the complex impact of weather conditions on driving risk, simplistically categorizing the influence of road conditions ("road") and weather conditions ("environment") into one category. This approach underestimates the extent of the impact of weather conditions ondriving risk and exhibits insufficient sensitivity to the risk calculation associated with extreme weather conditions, thereby significantly limiting the practical application of the method. Therefore, based on the DSF theory, a new environmental field function is introduced to achieve comprehensive coverage of risk factors in a"driver-vehicle-road-environment"framework. Specifically, the Behavior field, Kinetic energy field, Potential energy field, andEnvironmental field are constructed separately, and the BKPE model for driving safety field under adverse weatherconditions is proposed. In this study, the relevant parameters of the original driving safety field are re-calibratedbased on the Chinese road traffic safety dataset. Meanwhile, the exponential change characteristics of weather factors on driving safety are analyzed, and an environmental impact factor is constructed, leading to the proposal of theenvironmental field function. On the basis of the driving safety field model incorporating the environmental field, the artificial potential energy function is calculated for specific cases using the Car-100 data set for micro-analysis.Two typical events are analyzed to quantify multiple types of risks, and a comparative analysis with the originaldriving safety field model is conducted, demonstrating that the original model underestimates the risks associatedwith weather conditions. Subsequently, based on Bootstrap sampling, the average accuracy rate of the artificial potential energy function in describing actual traffic events, calculated from six samples, reaches 91.7%. Finally, corresponding driving risk control strategies are proposed based on the BKPE model.

2025 Vol. 43, No. 2