Smart Transportation and Green Mobility Safety

Urgent need for traffic first aid and life rescue after a traffic accident. However, for a long time, there have been outstanding problems in the life rescue of traffic accidents, such as “delayed alarm, delayed rescue, and careless health care”. Discussion on life rescue system of a road transport accident. Aiming at minimize traffic accident deaths and injuries, and implement the strategy of Safe China and Healthy China. Based on traffic accident big data and proportional hazards modeling method. Incorporating Accident Trauma Lifecycle Timelines. Constructing the Formal Model of Life Rescue System in Road Transport Accidents. Establish a three-level life rescue system model for traffic accidents in line with national conditions. The actual effect is basically the same as expected. According to the model variable parameters, the key technology of the system is proposed. This provides guidance for practical application. The traffic accident life rescue system is based on human life, including information sharing, intelligent decision-making, and coordinated rescue, forming an integrated intelligent three-level life rescue system. This is effective in controlling traffic fatalities and disabilities.

This article investigates the detection problem of false data injection attack (FDIA) in intelligent transportation system. As the typical malicious attacks, the emergency of false data injection attack causes enormous challenge to the security of intelligent transportation system. To face this challenge, a detection ideal using state observer for FDIA is proposed. Through the established dynamic model of intelligent transportation system, the deceptive characteristic of FDIA is given. In contrast to the Chi-square detector, the detection detector using observer can capture the real-time state change caused by attack. Then, the proposed detection criteria using state residuals is given. At last, simulation experiments are presented to verify the effectiveness of the proposed detection method against FDIA on the intelligent transportation system.

In order to analyze the movement intention of pedestrians in traffic roads, this paper proposes a pedestrian tracking and predicting method incorporating fused attention. The method uses CenterTrack to track pedestrians, adds fused attention to its DLA34 to enhance the extraction of key features of the tracked objects and improve the efficiency of information transfer in the network, representing the extracted attributes of the tracked objects as points on a heat map and correlates them with a greedy matching strategy to track the objects accurately, then uses the LSTM to loop through the historical locations of the tracked objects to predict their future locations, and finally visualizes the tracked and predicted locations. The experiments show that the improved CenterTrack outperforms other tracking algorithms in many aspects, with MOTA reaching 91.9 and MOTP reaching 80.3; the predicted coordinate errors of LSTM for pedestrians are within a small range, with the highest MAE of 6.13px and the highest RMSE of 8.26px in 1280 × 720 video; combined with the visualization results. The experimental data and visualization results demonstrate that this method can effectively accomplish tracking and predicting the location of pedestrians in road conditions.

To improve the prediction accuracy of urban traffic flow under the emergency traffic accident, according to the characteristics of traffic accidents, put forward a kind of classification based on support vector regression (SVR) forecasting model, considering the characteristics of accident impact the change trend of traffic flow, according to the severity of the accident emergency traffic accident can be divided into minor impact, major impact and serious impact, and the relative delay time of actual accidents is used for cluster analysis to verify the validity of the classification method. Then, support vector machine is used to build different prediction models for the three accident categories, and the traffic flow under the corresponding categories is predicted. The validity of the proposed method is verified by analyzing the traffic data on the viaduct in Yan’an City. The results show that compared with the traditional SVR model, the prediction results based on the SVR classification regression method are more consistent with the actual data, and the traffic flow prediction errors under three types of accidents are 7.5959%, 9.9542% and 8.5704%, respectively. Therefore, the prediction accuracy of SVR classification regression model is better than that of traditional SVR model, and it is a more suitable and reliable method for traffic flow prediction in the case of urban traffic accidents.

Aiming at the difficult problem of lane line detection for autonomous vehicles under shadow conditions, a lane line detection method combining illumination-independent graph and improved Canny edge detection is proposed. According to the principle of light-independent map, the logarithmic chromaticity ratio space projection is performed on all pixels on the color image to obtain a light-independent map, which eliminates the interference of light and shadow; the maximum inter-class variance method is improved to segment the lanes Line and road background; use Hough transform to fit lane lines. The experimental results show that the method can accurately detect the lane lines under shadow conditions, and can meet the real-time requirements.

The subway tunnel clearance is directly related to the safe operation of trains and the maintenance of facilities. However, the low light of the tunnel causes the low accuracy and short inspection distance based on the video inspection method, which hinders the development of tunnel inspection. Aiming at the problems, this paper proposes an intrusion detection method for subway tunnel based on lidar, which uses lidar to scan long-distance subway tunnel scenes to obtain tunnel point cloud. Firstly, the tunnel section point cloud is obtained by preprocessing. Secondly, a three-dimensional tunnel bounding region is constructed by extracting and projecting the optimal bounding box. Finally, an adaptive matching mechanism between the bounding region and the tunnel point cloud is constructed to realize the judgment of tunnel intrusion. The algorithm proposed in this paper achieves 100% detection accuracy within 100 m of the tunnel, improves the detection accuracy of tunnel intrusion detection, and promotes the development of tunnel inspection.

Connected and autonomous vehicles (CAVs) employ advanced communication and automation control technologies that can drive more safely than (HVs) human-driven vehicles, thereby potentially enhancing traffic efficiency. In order to give full play to the advantages of CAVs, the construction of dedicated lanes（DL） has been regarded as an effective way. However, In the early stage of the development of CAVs, the construction of dedicated lanes for CAVs on freeway may lead to waste of road resources and require huge costs. Therefore, considering allowing CAVs to enter the existing HOV lanes, an innovative idea of shared lanes between CAVs and HOVs was proposed, and the traffic flow performance of the shared lanes was analyzed. Numerical simulations of the proposed four shared-lane management strategies were carried out by building a cellular automaton model of mixed traffic flow of two-way eight-lane freeway: SL (single-lane management strategy), (TCFC) vehicles under two lanes can change lanes freely management, (TCNC) vehicles under two lanes cannot change lanes, (TOAC) only CAVs under two lanes can freely change lanes. Parameters such as (HOVs and CAVs ratio) Pah, (CAVs penetration rate) Pa, flow, and speed were quantitatively analyzed to evaluate the adaptability of each shared lane management strategy. Simulation results show: with the penetration rate of CAVs increase, the traffic efficiency is improved. When Pah is 10–30%, it is recommended to adopt SL strategy, when Pah is 40–60% and Pa < 40%, adopt TCNC or TCFC strategy, and when Pa ≥ 40%, adopt TOAC management strategy. When Pah > 70%, it is no longer appropriate to set the shared lane, which will increase the congestion of the shared lane. The result provide policy recommendations for future freeway managers. It will help to improve lane utilization and stimulate people's use of CAVs and HOVs and improve road traffic efficiency.

Multiple roadside cameras sense and localize autonomous vehicles for Automated Valet Parking. The accurate multi-camera registration into a shared coordinate system is of great importance. Owing to the sparse and distributed placement of multiple cameras, there is a very small field of view (FOV) between the cameras. The multi-camera registration is complex and dependent on human expertise. Therefore, a multi-camera registration method with small FOV based on pose graph optimization is proposed in this paper to automatically generate the extrinsic. The moving checkerboard in the public FOV is utilized as a reference for calibration. The optimization method constructs a pose graph to store camera and checkerboard pose and solves the pose graph by calculating checkerboard reprojection errors. To increase the overall flatness of the ground, ground plane assumptions can be utilized to further optimize the parameters. The experimental results show that the proposed method has greater calibration accuracy on the constructed multiple RGB-D camera system, as well as being simple and efficient in the real calibration procedure.

In response to the diversification of driving travel demand, this paper establishes a two-layer route guidance system from the perspective of travel cost at the scenario layer and the execution layer, and further constructs a route selection model for multiple objectives in the mobile connected information environment. The corresponding fuzzy network model inference system is constructed and trained on the basis of this route selection model. The training results demonstrate the practical value of the model, enabling drivers to trust the best route provided by the system when traveling.

The application of big data methods is currently a hot topic in the field of road traffic accident forecasting. It is of great practical significance to clarify the current research progress and prospects. This article first summarizes the big data methods used in road traffic accident prediction, and then introduces the application of big data analysis in road traffic accident prediction problems, including accident frequency prediction, accident severity prediction, accident risk probability prediction, and the four major categories of accident impact prediction are analyzed, and their specific research scenarios are analyzed. Finally, it discusses several challenges faced by the use of big data methods in the field of road traffic accident prediction problems.

With the increasing quantity of vehicle drivers, traffic safety issues are becoming more prominent. Hazard prediction ability, driving operation ability, stress response ability, and rule recognition ability are important driving abilities and collectively influence the safety of drivers, so testing and evaluating their multiple abilities is essential. This paper designed typical driving scenarios, selected appropriate metrics for characterizing various driving abilities, and conducted simulated driving experiments to collect data. A personalized and differentiated evaluation model for drivers’ multiple abilities is constructed based on multiple layer data envelopment analysis. The results show that the simulated driving scenarios designed in this study can be used for driver safety testing, and the correlation analysis validates the necessity of performing this multi-level evaluation of driving safety. The proposed evaluation method has theoretical guidance and practical application value for driver screening, and the evaluation results can be used to generate personalized and targeted improvement recommendations.

The focus of intercity transportation research has shifted to integrated three-dimensional transport networks made up of many modes of transportation, including road, rail, and air. This study establishes a topological model of integrated three-dimensional transportation network for various components of the transportation infrastructure of an integrated three-dimensional transportation network to evaluate the connectivity of the transportation network thoroughly on the basis of existing single-mode transportation network connectivity indicators. It also creates an evaluation index system characterizing network connectivity from a static perspective. The integrated three-dimensional transport network’s connectivity is assessed at four different levels—node, transport hub, region, and network—and is verified using arithmetic examples. The approach suggested in this study can offer extensive scientific and intuitive decision-making support for the future development of multimodal transport infrastructure.

In order to improve the recognition accuracy of motor vehicle traffic conflicts at urban intersections, the conflicts between motor vehicles at intersections were recognized to establish a new traffic conflict recognition model based on the high precision trajectory data from the radar and video all-in-one machine at holographic intersections and in combination with complete vehicle motion information. Conflict categories were identified; and the method for determining the severity of vehicle traffic conflicts was provided. Points and severity of conflicts at an intersection in Beijing were recognized. The quantitative description of potential safety risks at this intersection was identified.

In order to assess the operation risk of freeway operating buses, a method of operating risk assessment of freeway operating buses based on GPS data is proposed by using the interval type-2 triangle fuzzy theory. The frequency of continuous deceleration of operating buses, average speed and freeway horizontal and longitudinal design parameters were selected as risk assessment indexes, and operation risks were divided into four levels: safe, safer, more dangerous and hazardous. The upper and lower membership functions of the interval type-2 triangular fuzzy number were constructed by using the boundary limits of each risk level, and the corresponding interval membership matrix was constructed according to the value of the evaluation index. The index weight value was calculated by using the overrate method, and the interval membership matrix and index were weighted and summed to get the comprehensive membership interval. Finally, the membership vector of the risk level was determined by using the possibility ranking method. The risk level is evaluated according to the principle of maximum membership degree. The results show that the evaluation method has realized the quantitative and graded evaluation of the operating risk of operating buses on freeways, and identified the sections with poor safety. Taking the G12 Hui Wu Freeway from Jilin to Changchun Longjia Airport as an example, the effectiveness of the evaluation method has been verified.

To accurately assess the impact of station failure on the transportation capacity of the high-speed rail (HSR) network under gale conditions, an improved space-L method was used to construct a directed weighted network topology model based on high-speed train frequencies. A comprehensive vulnerability evaluation index for the HSR network under train frequency conditions was then proposed. Next, based on the improved Coupled Map Lattice (CML) model simulates the cascading failure process of the network caused by the failure of network nodes under different external disturbance R and coupling coefficients, and measures the changes in network vulnerability during the cascading failure process. Finally, the validity of the model was verified and the vulnerability of the 2022 HSR network was analyzed as an example. The research findings indicate that the existing HSR network in China exhibits scale-free and small-world network characteristics. As the external disturbance R increases, the range of cascading failure in the network expands, leading to an increase in the resulting fragility. At different coupling strengths, the range of cascading failures in the ε1 is more destructive to the network than the ε2. Therefore, it is important to focus on protecting stations that are vulnerable to gale damage and preventing faults from spreading within the network.

Bad weather can negatively affect the normal operation of highways, and prediction of traffic dynamics under bad weather is an effective means to improve the efficiency of highway traffic and enhance safe operation under adverse weather conditions. To solve the above problems, this study establishes a DBN-AdaBoost prediction model based on highway weather data and traffic flow data. Firstly, the DBN model is used to extract the effective features of weather data, and the weights and biases of DBN are optimized continuously and iteratively. Then the BP-AdaBoost traffic pattern prediction model is constructed based on the effective features. Finally, the actual highway traffic flow data is selected for validation. The results show that the mean square error, mean absolute error, and mean square percentage error of the proposed DBN-AdaBoost prediction model are lower than those of other prediction models, and the prediction error is the smallest and the accuracy is the highest, which can complete the prediction of highway operation situation under severe weather.

The construction of freeway has brought remarkable economic and social benefits. However, with the growth of traffic demands, freeway traffic incidents occur more and more frequently. The freeway traffic incidents will bring significant inconvenience to the road users. For the highway management administration, the appearance of the incident will disturb the normal operation order of the highway, it is necessary to take timely measures to reduce the impact of abnormal events. In this study, traffic flow status data such as freeway incident data, traffic flow, speed and occupancy were used to develop a Random Forest-Support Vector Machine (RF-SVM) method for incident recognition. The results show that the model is feasible to predict the risk of traffic incident.

To compensate for the fixed-parameter model's tendency to lead to biased estimates and erroneous inferences of results, a sample of 2016 car-vehicle traffic accident data in France was used, with accident severity as the dependent variable and 12 factors such as people, roads, environment, and accident characteristics as independent variables, based on a random parameter logit model using a backward stepwise selection method, while using average marginal effects. The results showed that: ① the variables “Passenger ≥ 2”, “Curve road”, “One-way road”, “Two-way physically separated road, “No public lighting at night”, “Severe weather”, “Urban area”, “head-on collision “ as random variables; ② the random parameter Logit model fits better compared to the ordinary logit model, which can effectively capture the unobserved heterogeneity in accident data and is more suitable for traffic accident analysis.

Using the interweaving zone road section of US101 in NGSIM as a research case, the effects of various data processing methods were compared, and the characteristics of lane-changing behavior in the interweaving zone were statistically analyzed, and it was found that most of the lane-changing behavior on the detected road section of US101 occurred in the middle section. The results demonstrate that the Multi-head CNN-BiLSTM model has higher accuracy and stronger robustness than the LSTM model in recognizing lane-change intentions in the weaving zone of the highway. The research results can be used in connected autonomous vehicles and vehicle–road collaboration systems for recognition judgments.

Accurate identification of driver’s distraction is of great significance to prevent traffic accidents. In this study, the information of driver’s body joint points is fully utilized, and a machine vision-based biomechanical distraction identification method is proposed. A real vehicle driving experiment is designed and organized. The normal driving data and biomechanical distraction data of 20 participants are collected to construct a data set. The Lightweight OpenPose network is trained to extract the position information of the driver’s joint points. A number of characteristic parameters such as the driver’s limb angles or Euclidean distance between the joint points can be calculated based on the position information. The experimental data are preprocessed using moving average filter and factor analysis. The biomechanical distraction identification model is trained with Particle Swarm Optimization (PSO) and Probabilistic Neural Network (PNN), and an accuracy rate of 91.5% is achieved. The possibility of using body joint points to identify the biomechanical distraction is demonstrated in this study. It provides theoretical and technical support for the identification and application of driver’s biomechanical distraction. It is of great significance to improve the active safety performance of smart vehicles.

At present, although China’s traffic accident rate has decreased year by year, but people have been shocking for the huge number of casualties and property losses. Safety is an important issue in the transportation field. In the future, as the technology of Internet of vehicles becomes mature, the construction of road intelligent infrastructure follows up steadily, and the breakthrough of realizing autonomous driving, the beautiful vision of vehicle–road collaboration will be realized. At that time, the safe requirements for vehicle driving will be more stringent, and the correct driving decision is the premise of realizing safe driving. Therefore, this paper aims to provide drivers with more accurate and timely lane-changing decision intervention and construct a lane- changing decision model considering driving style. First, according to the lane-changing decision definition to divide the vehicle trajectory, and then the lane-changing decision characteristics are selected according to the vehicle interaction relationship. In this paper, the channel change decision model is constructed based on NGSIM data. Therefore, the two-step trajectory reconstruction technology is used to deal with data outliers and noise and extract relelant features, Basing on k-means clustering method, driving style calibration is achieved using following stage data. Finally, the XGBoost lane change decision model was constructed, and the necessity of adding driving style features and the superiority of XGBoost compared with RF and SVM models were verified through comparative experiments.

One of the main challenges in limiting the application of conditional driving automation is the presence of safety hazards during the process of drivers taking over vehicle control. In order to deal with the takeover safety challenge, automated driving system can provide the driver with safety control assistance under some complex and dangerous takeover conditions, and directly hand over control to the driver only under some simple and safe takeover conditions. There are no other vehicles around the ego vehicle and the weather is good, which is a relatively simple takeover scenario. In this scenario, takeover time budget and non-driving related tasks (NDRTs) are the key to determine whether the automated driving system can directly hand over control to the driver. This paper designed a conditional automated driving takeover experiment based on a driving simulator, recruiting a total of 45 participants to explore the impact of takeover time budget (5 s, 7 s, 10 s) and NDRTs (monitoring driving, non-visual NDRT, visual NDRT) on drivers’ takeover performance in the simple takeover scenario. Statistical analysis showed that both the takeover time budget and NDRTs significantly affected the driver’s takeover time and quality. More importantly, we found that under the condition of 5 s takeover time budget, the driver's visual distraction state could cause a near-crash situations (minimum TTC after takeover < 1 s). This result indicates that when the driver is visually distracted under 5 s budget condition, the system should also provide safety control assistance in the simple takeover scenario. When the takeover time budget is greater than 5 s, the system can directly hand over control to the driver. This conclusion can provide a theoretical basis for the safety design of conditional automated driving system.

The tunnel is a closed, space-tight structure, easy to affect the safety of the driver, and the severity of traffic accidents inside the tunnel is relatively higher. In order to explore the differences in driving behavior in different length tunnel environments on expressway, and promote the tunnel targeted safe operation, 10 experienced drivers are recruited to carry out the real vehicle test in the Zhongnanshan tunnel group of Qinling Mountains. Capture vehicle steering data, speed data, and driver heart rate data using steering wheel goniometer, CAN-OBD, and heart rate monitor. After data pre-processing, the vehicle steering characteristics, speed characteristics and driver heart rate characteristics are analyzed in time domain. Vehicle steering angular velocity, vehicle speed and driver's heart rate signal are converted into the frequency domain by Fourier transform for spectrum analysis. The results show that when the driver is driving inside the tunnel, as the length of the tunnel increases, the speed of the steering angle decreases first and then rises, the vehicle speed is on the downward trend, the heart rate and heart rate growth rate of the driver are on the upward trend. In the meanwhile, the longer the tunnel length, the lower the subjective active participation of the driver is, the worse the vehicle's speed retention capability is, and the heart rate stability also gradually deteriorates. The manual simulation environment in the extra-long tunnel can create psychophysiological stimulation for the driver, relieve the load of long-term driving in the tunnel, and improve the stability of driving performance and driving behavior indicators. The research findings can provide theoretical guidance for developing and improving the safety policy of different length tunnels.

In order to study the severity of highway accidents and deeply explore their influencing factors. Collected typical highway accident data and traffic flow data in Henan Province, constructed Logit and mixed Logit models based on single vehicle accidents and multi vehicle accidents, and selected the optimal model; Establish random forest and Gradient Boosting Decision Trees, build a prediction model of expressway accident severity, screen the characteristics of expressway accident severity to obtain the importance ranking of influencing factors, and select the machine learning algorithm suitable for this study through the model prediction accuracy test; Compare and analyze the two optimal models using the K-fold cross test method, and analyze the core influencing factors selected from the optimal models based on the backward elimination. The results indicate that the mixed Logit model to some extent solves the limitations of the Logit model in analyzing factors affecting the severity of accidents; Machine learning algorithms have higher prediction accuracy.

The weaving area is a potential bottleneck that affects the efficiency and safety of freeways and urban expressways. As a typical hub interchange, the weaving section inside the cloverleaf interchange is generally short, and the probability of vehicle collision accidents is also higher. Therefore, this paper profoundly analyzed the conflict risk of the weaving areas of cloverleaf interchanges and its influencing factors. A methodological framework based on UAV videos was constructed to identify and track vehicles and extract high-precision trajectories. Based on trajectory data, an extended time-to-collision was proposed to identify traffic conflicts and evaluate collision risks. Subsequently, the spatial distributions of traffic conflicts were examined, and the risks under different operating conditions and traffic states were discussed. Finally, a regression model was established to quantify the impact of traffic variables on conflict risks. The results show that the proportion of lateral conflicts in the cloverleaf interchange weaving area is higher and more severe than that of longitudinal conflicts. The concentrated region of traffic conflicts is the merging area at the front of the weaving area, and it presents an exponential attenuation trend with the segments of the weaving area. The conflict risk under different operating conditions and traffic states varies significantly. The conflict risks from mainline-to-ramp vehicles are higher than those from ramp-to-mainline vehicles and mainline-to-mainline vehicles. Moreover, the more congested the traffic state is, the higher the potential collision risk in the weaving area. These results can help traffic participants develop reasonable risk control plans for weaving areas.

In order to find out the road sections that may have traffic safety hazards in the design stage of expressway, this study carried out the design of expressway safety evaluation system for the design stage. Based on the analysis of system requirements, requirements objectives and workflow, the safety evaluation system of highway design stage is designed, including database module, specification compliance review module, design consistency evaluation module, accident prediction module, map operation module and user operation module. In this study, a safety evaluation model based on design consistency and accident prediction factors is presented, utilizing the integration of MapInfo and SQL Server database. The model library is constructed, and subsequently, standard compliance review and design consistency evaluation are conducted. The system can realize visualization and docking with road design software. The adjustment of design scheme and data statistics and analysis can be realized by using the callback technology of MapInfo. The design, evaluation and optimization of expressway can be coordinated synchronously, and the safety of expressway in design and operation stage can be evaluated.

In order to realize the safety evaluation of freeway alignment design in the design and construction stage and find out the potential traffic safety hazards, carry out a study on the expressway traffic accident prediction model. Based on the relevant alignment design, traffic characteristics and accidents data, build the primary accident prediction model, considered different conditions of alignment design and traffic control features of the influence of the traffic accident, adjusted the primary accident model by accident correction factor, established an active expressway accident prediction model based on alignment design and traffic characteristics, and carried out the goodness of fit test and example verification. The results show that the fitting effect of the model is good, and the accident prediction results are close to the actual accident data, which proves the validity of the model. The research results can provide theoretical basis for freeway improvement measures in the design and construction stage and improve the level of freeway traffic safety.

The vehicle sound insulation and high-speed driving cause Doppler effect of roadside auditory warning audio, which greatly weakens drivers’ auditory perception and sensitivity, and affects the warning effect. High-frequency audio refers to the frequency close to the limit of human natural sound. In order to explore the necessity and feasibility of applying high-frequency audio to roadside auditory early warning, the finite element method was used to simulate the transmission process of roadside warning audio during vehicle driving. Compared the audio warning effects of 12 working conditions, including 6 frequencies, and window opening or closing, and the sound pressure level (SPL) was used as an indicator to measure the driver’s auditory perception. Meanwhile, taking Jiashao Bridge as the actual project case study, the high-frequency audio design method suitable for roadside warning was proposed. The results show that the frequency of audio is increased from 250 to 2000 Hz, the SPL is increased by about 20 dB, and the auditory perception of audio is obviously enhanced. High-frequency audio (frequency between 1000 and 1500 Hz) can reduce the impact of vehicle sound insulation. The results effectively verify that High-frequency audio can improve the auditory perception of human ear, which has important significance for improving the effect of roadside warning and ensuring driving safety.

Human driving skills and perception are replaced by programs in Autonomous vehicles (AVs). However, existing road infrastructures were not designed with AVs in mind, leading researchers to examine their feasibility on these roads. This study uses virtual simulations to answer two key questions: how AVs perceive combined alignments and how they navigate on them. The study uses PreScan and MATLAB/Simulink to simulate real-world scenarios for a Level 4 (L4) AV. The first question explores the relationship between AV operating speed (Vo) and lane departure to understand AV perception. The second question investigates AV driving trajectories on combined road alignments. Results show that the point where tangent meets the spiral curve significantly affects AV perception. Additionally, left turns exhibit more deviation than right turns, with AVs showing significant left deviation behavior before entering a curve. These findings offer insights into AVs’ performance on complex combined alignments, aiding the development of AV technology for safe and efficient operation on existing roads.

Aiming at the problems of slow detection speed, large prediction error of warning area and weak environmental adaptability of the current machine vision-based vehicle collision warning technology, this paper proposes a collision warning system based on the recognition of slippery road conditions. Firstly, this paper uses the on-board camera to monitor the environment and road conditions in front of the vehicle in real time, and uses the YOLOv5 algorithm to detect the vehicle in front of it in real time, while accurately identifying the current wet state of the road, such as dry and slippery, through the ResNet50 model in the convolutional neural network. Secondly, a driving safety distance model with adaptive traffic environment characteristics is established by combining different road environments and driving conditions, and an early warning area is generated that changes dynamically with the speed of the vehicle and the slippery state of the road. Finally, possible collisions are predicted and warned in time, based on the relationship between the area of the warning and the position of the vehicle. Experimental results show that the method proposed in this paper improves the overall warning accuracy by 6.72% and reduces the warning false alarm rate for oncoming traffic on both sides by 16.67% compared with the traditional risk warning algorithm. Its application in practical driving can effectively ensure the safety of the driver and has a high application value.

During the driving process, drivers’ misjudgment and improper operation are highly likely to lead to traffic accidents. This article aims to explore the mechanism of dangerous car following behavior and construct a predictive model for dangerous car following behavior that considers multiple factors such as environment and vehicle interaction. First, we propose a car following driving behavior extraction method based on trajectory data, establish a three-dimensional driving behavior feature set of vehicle motion characteristics, vehicle interaction characteristics and microscopic traffic flow characteristics, and use random forest model to screen key features. Subsequently, the K-means algorithm was used to optimize the unbalanced dataset of dangerous car following behavior, and a dangerous car following behavior prediction model based on K-GMMHMM was proposed, and the prediction accuracy was compared with other models. The prediction performance of K-GMMHMM was found to be better, with an accuracy rate of 97.86%, verifying the effectiveness of the proposed method for predicting dangerous car following behavior. This provides theoretical support for the development and application of vehicle warning assistance systems, and has practical significance and application value for improving road traffic safety and preventing traffic accident risks.

With the promotion and application of technologies such as the Internet of Things, big data, and artificial intelligence in the transportation field, intelligent vehicle road collaborative systems have become an important means to improve the efficiency of transportation systems. This study aims to leverage the characteristics of autonomous vehicles and propose a collaborative traffic control strategy for autonomous vehicle intersections based on preemptive level, in order to provide reference for the improvement of safety and efficiency in future intersection operations. According to the intersection environment information, generate the track routes of automatic driving vehicles at different exit lanes to pass through the intersection, study the coupling space-time constraints of different track routes at the intersection, judge whether there is conflict according to the minimum safe headway of vehicle traffic, and obtain the conflicting point sequence; Calculate the preemptive level based on the time it takes for autonomous vehicles to reach each conflict point on each trajectory route, determine the vehicle traffic order based on the preemptive level, and resolve conflicts one by one for the conflict points that exist. Finally, select trajectory routes facing multiple exits based on traffic efficiency; Using Sumo and Python to build intersection scenarios for simulation, under high and low input flow conditions, the simulation results of collaborative traffic control for autonomous vehicle intersections based on preemptive level are significantly better than those without control strategies in the seven evaluation indicators determined in the dimensions of safety, green, and high efficiency. The average travel time has increased by 71.52% and 63.66%, the average standard deviation of speed has decreased by 60.84% and 43.37%, and the average fuel consumption decreased by 56.20% and 37.16% respectively, verifying the effectiveness of the motion optimization strategy.

Trajectory data serves as the foundation for scientific research in traffic engineering, providing valuable insights for studies on traffic safety and the development of highly autonomous driving technologies. This paper introduces the CQSkyEyeX dataset, a comprehensive and large-scale collection of natural vehicle trajectory data from various scenes on China's expressways. The technical system of extracting trajectory data from high-resolution aerial video is also outlined. The dataset comprises 600 min of measurements from eight expressway locations, encompassing expressways’ basic sections, weaving segments, and merge/diverge segments. It encompasses rich road infrastructure information and a wealth of trajectory data indicators. To ensure accuracy, the trajectory data was validated using data acquisition instruments and manual measurements. The obtained results indicate trajectory position errors of typically less than 10 cm, velocity errors below 1.5 km/h, and evaluation indices for multi-target tracking, with MOTA and MOTP scores reaching 97.2% and 96.8%, respectively. By incorporating these comprehensive and fine-grained vehicle trajectory data, our dataset enables a more microcosmic analysis of vehicle behavior, facilitating in-depth investigations into traffic safety and related research areas. CQSkyEyeX dataset can be obtained online for free at www.cqskyeyex.com and continuously updated.

This article studies traffic target detection and parameter extraction from the perspective of drones. A YOLOv7-based improved traffic object detection algorithm and traffic parameter extraction method are proposed to address the problems of low detection accuracy for dense small targets and large extraction errors in traffic object detection tasks from the perspective of unmanned aerial vehicles. Add an attention mechanism after the original Backbone network to improve the feature extraction efficiency of the model; The EIoU position loss function is used to optimize the update efficiency of parameters. Maintaining a balance between detection speed and accuracy on the VisDrone dataset, the improved algorithm achieved a 2.20% increase in map compared to YOLOv7, demonstrating the effectiveness of the object detection algorithm and improving the extraction of traffic parameters in this scenario compared to general methods.

Work zone poses great threaten to traffic safety and efficiency. One important reason is the frequent lane-changing behaviors. This study carried out a simulation study to investigate driver’s merging behavior in depth and verify the efficacy of an intelligent in-vehicle audio prompt. Forty-two participants completed the test and both driving behavior and eye movement data were collected. During analysis, a multistage behavioral mechanism model was proposed including perception stage, preparation stage and action stage. The results showed that: (i) Professional drivers had better control over the longitudinal and lateral movement during lane changing; (ii) High traffic density conditions would postpone lane changing and led to aggressive driving behavior. Besides, drivers under high traffic density conditions had less concentration on the road ahead, which might increase the collision risks. (iii) The audio prompt could induce lane changing more in advance and provide more time for lane changing preparation. Moreover, the audio prompt could also reduce drivers’ visit duration in rearview mirror in preparation stage, thus ensuring driving safety. This research can shed some lights on traffic management in work zone areas.

Objective: This paper aims to explore the potential relationship between driving anger and personality traits concerning with the hazmat truck drivers, and then provides a scientific basis for effective psychological intervention strategies and driver safety training. The ultimate goal is to minimize transportation risk caused by driving anger and improve the safe driving level of drivers. Subjects and Methods: Firstly, based on the actual investigation, the typical driving scenes of hazmat truck drivers and the factors that induced their anger state were determined, and a Driving Anger Scale (DAS) specified for them was designed. Secondly, 555 professional hazmat truck drivers were recruited to complete DAS and NEO-five Factor Inventory (NEO-FFI). On this basis, the DAS revised by using item analysis, Exploratory Factor Analysis (EFA), Confirmatory Factor Analysis (CFA) and reliability and validity analysis. Finally, the relationship model between personality traits and driving anger of hazmat truck drivers was proposed, verified and modified by using the Structural Equation Model (SEM). Results: The revised Driving Anger Scale could be divided into four dimensions, including illegal behaviour, discourtesy, traffic obstruction and transport interference, and had adequate fit indices, convergence and discriminant validity. The results of SEM showed that there was a positive relationship between neuroticism and transport interference (β = 0.593, t = 10.958, p < 0.001), illegal behaviour (β = 0.652, t = 8.754, p < 0.001), traffic obstruction (β = 0.837, t = 10.319, p < 0.001) and discourtesy (β = 0.649, t = 10.846, p < 0.001) subscale score on the Driving Anger Scale. Agreeableness (β = 0.341, t = 4.521, p < 0.001) had a significant positive effect on traffic obstruction. Both conscientiousness (β = 0.141, t = 1.976, p < 0.05) and extraversion (β = 0.155, t = 2.011, p < 0.05) had significant positive effects on illegal behaviour subscale score. Conclusions: Hazmat truck drivers with high levels of neuroticism were prone to experience more driving anger than those with low levels of neuroticism. Those with high agreeableness would be angry with traffic obstruction on the road. And those with high extraversion and conscientiousness were prone to anger at illegal behaviour. But openness bad a negligible relationship with driving anger. The results of this study provided evidence for the use of personality traits of hazmat truck drivers to predict their level of driving anger. It provided a basis for more scientific driver anger measure, and a better application to driver anger interventions and psychological health education for hazmat truck drivers.

The purpose of this study is to explore the variations in the dynamic performance of trucks with different loads and initial velocities on different slopes. The Trucksim simulation software is used to simulate a three-axle tractor and a two-axle heavy low-bed trailer with a road adhesion coefficient of 0.85, and a large loaded truck climbing dynamics simulation model is constructed. A comparative analysis was conducted on the speed variations of the large loaded truck under different conditions, including a design speed of 120 km/h and longitudinal slope gradients of 4 and 5%, as well as a design speed of 100 km/h and longitudinal slope gradients of 3, 4, and 5%, under empty, half-loaded, and full-load conditions, respectively. The influence of various factors on the steady speed reached by the truck during climbing was studied using the controlled variable method. The results showed that the initial climbing speed, truck load, and slope gradient have a significant impact on the steady speed of the truck after climbing. For mountainous highways with a slope gradient of 3%, there is no need to set climbing lanes. However, when the slope gradient exceeds 3%, the probability of a speed difference of over 20 km/h before and after climbing increases, suggesting the need for climbing lanes to be set.

This chapter presents a concise review of the current research status in traffic flow data collection, estimation, prediction, and state estimation. It summarizes the main research trends, and achievements, and identifies the associated issues and limitations. Regarding data collection methods, vehicle-mounted GPS, microwave detection, video detection, and highway toll station data are discussed. GPS data has limitations due to its reliance on vehicle coverage and interruptions caused by signal quality, limiting its representation of overall travel information. Hence, emphasizing the practicality and importance of using limited traffic flow data for road segment flow prediction. Flow prediction commonly employs parametric statistical models and non-parametric machine learning models, with machine learning models exhibiting better estimation accuracy. Further exploration is needed for traffic flow prediction using GPS data. In terms of traffic flow prediction, recent research focuses on utilizing historical data combined with non-parametric machine learning or deep learning models. Exploring traffic flow prediction based on distinct operational features is crucial for understanding traffic on road segments. The analysis of state estimation methods encompasses traffic feature indicators, estimation models, complexity, and classification. Existing studies primarily concentrate on single vehicle types, with limited exploration of parameters for passenger cars and trucks. Additionally, there are differences in determination criteria and levels among government agencies. To predict traffic state accurately with limited parameters, it is essential to explore the parameter ranges specific to different vehicle types.