No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
A dynamic method to predict driving risk on sharp curves using multi-source data
Abstract
Traffic accidents are likely to occur on sharp curves under poor driving conditions, and the severity level of such accidents is high. Therefore, predicting the risk associated with driving on curved roadways in real time can effectively improve driving safety. This paper aims to develop a dynamic real-time method that fuses multiple data sources to predict risk when driving on sharp curves in the context of the connected vehicle environment. Six curves with three small radii (60 m, 100 m, 150 m) and two driving directions (left and right) were designed for a driving simulation experiment. Driver maneuvering data, vehicle kinematic data, and physiological data of 55 drivers were collected for this study. The data were combined and spatially and dynamically segmented. The mean value of the critical lateral acceleration of the vehicle was set as the risk assessment index. K-means clustering was used to classify the driving risk into three levels: low, medium, and high. Then, the risk level was predicted using the maneuvering data, vehicle kinematic data, and physiological data as well as road alignment characteristics as input features for the proposed model that employs the long short-term memory (LSTM) network algorithm. Models with different combinations of observation window (lookback) and interval window (delay) were compared to derive the best window combination. The algorithms selected for comparison against the LSTM algorithm are random forest, XGBoost, and LightGBM. The results show that the proposed LSTM-based method can effectively predict dangerous driving behavior on sharp curves. The optimal window combination derived using the LSTM algorithm is lookback = 20 m and delay = 20 m. The prediction performance of the proposed model is significantly better than that of the other three compared algorithms, with F1-scores of 84.8% and 86.0% for the medium and high risk categories, respectively. In addition, the proposed LSTM-based model that fuses multiple data sources is proven to outperform the model that uses only vehicle kinematics data. The dynamic prediction method proposed in this paper can contribute to the development of a real-time prediction and warning system for driving risks at vehicle terminals in the intelligent connected vehicle environment.
... A deficiency in any of these abilities can lead to serious traffic accidents and heightened driving risks. Young drivers (aged [17][18][19][20][21][22][23][24][25] and older drivers (aged 60 and above) are particularly considered high-risk groups [10]. Younger drivers, especially those newly licensed, may encounter unfamiliar driving challenges and exhibit impulsivity, leading to relatively weaker cognitive driving abilities [11,12]. ...
... Most current studies on driving-related cognitive abilities assessments focus on specific cognitive components such as attention [15][16][17][18], reaction speed [19][20][21], working memory (WM) capacity [22][23][24], or perceptual ability [9,25,26]. For example, Zatmeh-Kanj and Toledo [16] proposed the Gipps' Model (GM) for car-following based on vehicle data, a classical model used to assess driving attention. ...
... Sensors 2025,25, 174 ...
With the increasing complexity of urban roads and rising traffic flow, traffic safety has become a critical societal concern. Current research primarily addresses drivers’ attention, reaction speed, and perceptual abilities, but comprehensive assessments of cognitive abilities in complex traffic environments are lacking. This study, grounded in cognitive science and neuropsychology, identifies and quantitatively evaluates ten cognitive components related to driving decision-making, execution, and psychological states by analyzing video footage of drivers’ actions. Physiological data (e.g., Electrocardiogram (ECG), Electrodermal Activity (EDA)) and non-physiological data (e.g., Eye Tracking (ET)) are collected from simulated driving scenarios. A dual-branch Transformer network model is developed to extract temporal features from multimodal data, integrating these features through a weight adjustment strategy to predict driving-related cognitive abilities. Experiments on a multimodal driving dataset from the Computational Physiology Laboratory at the University of Houston, USA, yield an Accuracy (ACC) of 0.9908 and an F1-score of 0.9832, confirming the model’s effectiveness. This method effectively combines scale measurements and driving behavior under secondary tasks to assess cognitive abilities, providing a novel approach for driving risk assessment and traffic safety strategy development.
... some studies have predicted the likelihood of a person engaging in aggressive driving behaviour by analysing self-reported data obtained from the aggressive Driving Behaviour Questionnaire (aDBQ) (Gurda, 2012). More recently, most studies have used more objective obtained driving data to identify and predict driving behaviour in real time (lee & Jang, 2019;Ma, Wang, et al., 2023;Wan et al., 2023). ...
Emotion is an important factor that can lead to the occurrence of aggressive driving. This paper proposes an association rule mining-based method for analysing contributing factors associated with aggressive driving behaviour among online car-hailing drivers. We collected drivers' emotion data in real time in a natural driving setting. The findings show that 29 of the top 50 association rules for aggressive driving are related to emotions, revealing a strong relationship between driver emotions and aggressive driving behaviour. The emotions of anger, surprised, happy and disgusted are frequently associated with aggressive driving behaviour. Negative emotions combined with other factors (for example, driving at high speeds and high acceleration rates and with no passengers in the vehicle) are more likely to lead to aggressive driving behaviour than negative emotions alone. The results of this study provide practical implications for the supervision and training of car-hailing drivers.
Microscopic traffic models serve as indispensable tools in tasks such as constructing test scenarios for autonomous vehicles (AVs), predicting trajectories, and analyzing traffic flow dynamics. However, a significant proportion of these models rely on assumptions of normal behaviors. Yet, the validity of these assumptions is dubious given the heterogeneous nature of traffic flow and existence of abnormal driving behaviors. These limitations impede the efficacy of conventional microscopic models in crucial tasks like constructing AV test scenarios with specified risk levels, analyzing abnormal behaviors, etc. To address these challenges, this study contributes by proposing a model tailored to accommodate two-dimensional abnormal driving behaviors in microscopic traffic framework. The proposed approach have the following innovations: 1) it incorporates assumptions concerning abnormal behaviors in both the longitudinal and lateral dimensions; 2) abnormality at each dimension is captured by a combination of certain terms; 3) stochastic control barrier method is applied to customize the risk levels of the resulting traffic flow dynamics. Additionally, we present a method for retrieving vehicular maneuver information, enabling the extraction of detailed vehicle body gestures and driver control inputs, which would benefit the analysis of abnormal behavior. Our findings demonstrate that the proposed model yields longitudinal and lateral dynamics consistent with empirical observations, and various abnormal behavior patterns can be simulated.
Fatigue and distraction are the most common long-term poor state and short-term abnormal behavior of drivers, significantly increasing the driving risk of vehicles equipped with the advanced driver assistance system (ADAS). To provide a more reliable decision-making basis for ADAS and improve driving safety, this paper proposes a driving risk assessment framework considering the driver’s long-term poor state and short-term abnormal behavior. Firstly, based on the self-built fatigue dataset and transfer learning method, an adaptive fatigue detection model with strong generalization capability is established to enable multi-view driver fatigue detection. Then, the idea of multi-clustering and adding offset parameters is introduced into the classical contrast loss function, and the D-InfoNCE loss function is designed to realize the accurate identification of the driver’s specific distraction behavior under open set detection. Subsequently, a driving risk assessment system is developed to quantify driving risk based on the vehicle driving risk factors when fatigued or distracted driving occurs. Finally, the proposed driving risk assessment system is validated by the datasets and driver-in-the-loop test bench. The results show that the proposed framework can accurately detect the driver’s fatigue state and distraction behavior and give ADAS the corresponding driving risk levels to enhance driving safety.
Safe traffic is an important part of sustainable transportation. Road traffic accidents lead to a large number of casualties and property losses every year. Current research mainly studies some types of traffic accidents and ignores other types of traffic accidents; therefore, taking various types of road traffic accidents as a whole, an overall study of their influencing factors is urgently needed. To improve road traffic safety, taking various types of road traffic accidents as a whole, this paper analyzes the influencing factors and finds out the causative factors of road traffic accidents. A new index system of road traffic accident influencing factors is constructed based on the existing literature and real traffic data, and their subjective weights and objective weights are obtained by the analytic hierarchy process based on the subjective data and the normalization of the actual traffic data for Yizheng City, Yangzhou, China from January 2020 to December 2020, where the subjective weights are the main weights, and comprehensive weights are obtained by the minimum discrimination information principle correcting the subjective weights with the objective weights. Finally, the global weights, their ranks, and their weight differences are obtained. The main findings are as follows: (1) compared with the real traffic data, experts generally overestimate the impact of road factors on traffic accidents and underestimate the impact of human factors on traffic accidents; (2) in the first-level, human factors and road factors are the causative factors; (3) in the second-level, “motor vehicle drivers’ misconduct”, “road condition”, and “road section” are the causative factors; and (4) in the third-level, “slippery road”, “rain and snow weather”, “intersection”, and “untimely braking” are the causative factors. The research results can provide some scientific basis for improving road traffic safety.
Reduced-scale mobile robots (RSMRs) are extensively used for studying autonomous driving due to their ability to test models and algorithms in physical environments, their lack of constraints related to regulations and laws, and their advantages of low cost and space-saving. Nevertheless, there is currently a lack of systematic analysis and review of these autonomous driving studies involving RSMRs. Hence, this paper comprehensively reviews 134 studies on the application of RSMRs in autonomous driving research. Through the analysis of these studies, we summarize the commonly used methods for three modules (perception, decision-making, and actuation) of the autonomous driving process of RSMRs, and thoroughly examine the main applications (navigation and obstacle avoidance, vehicle fleet coordination, intersection management, parking control, drift control, passenger unease, and hands-free control) covered in these studies. Furthermore, we identify the limitations and gaps in the existing studies related to RSMRs, and provide recommendations for future research initiatives: 1) focusing on common interactive driving events in real-world traffic such as lane changing, merging, cut-in, and overtaking, 2) extending the experiment duration and distance, 3) increasing the randomness in experimental design, 4) exploring the transferability of autonomous driving algorithms from RSMRs to real vehicles, 5) researching on the mixed fleet consisting of manually controlled RSMRs and self-driving RSMRs.
Virtual reality (VR) driving simulators are very promising tools for driver assessment since they provide a controlled and adaptable setting for behavior analysis. At the same time, wearable sensor technology provides a well-suited and valuable approach to evaluating the behavior of drivers and their physiological or psychological state. This review paper investigates the potential of wearable sensors in VR driving simulators. Methods: A literature search was performed on four databases (Scopus, Web of Science, Science Direct, and IEEE Xplore) using appropriate search terms to retrieve scientific articles from a period of eleven years, from 2013 to 2023. Results: After removing duplicates and irrelevant papers, 44 studies were selected for analysis. Some important aspects were extracted and presented: the number of publications per year, countries of publication, the source of publications, study aims, characteristics of the participants, and types of wearable sensors. Moreover, an analysis and discussion of different aspects are provided. To improve car simulators that use virtual reality technologies and boost the effectiveness of particular driver training programs, data from the studies included in this systematic review and those scheduled for the upcoming years may be of interest.
The free-flowing traffic environment of the freeway is an important application scenario for automatic driving. In this scenario, the freeway’s geometric design is an important factor because no other vehicle affects the driving process of the target vehicle. The freeway’s combined curves have more safety problems, but there are no quantitative guidelines for their geometric design. They present more challenges for automatic driving or driver assistance functions. If the relationship between human-drivers’ micro-behavior and the geometric design of combined curves is examined, it could provide theoretical support for the enhancement of automated driving and driver assistance functions as well as the quantitative design of combined curves. The paper analyzed the speed change and lane departure behaviors of combined curves, considering downslope curves, upslope curves, sag curves, and crest curves. The relationship between micro-driving behaviors and combined curves’ geometric design were determined using random forest models. The SHAP values of each variable were calculated. The results showed that (1) on a downslope curve and sag curve the speed change behavior should be paid more attention; on an upslope curve and crest curve, the lane departure behavior should be paid more attention; (2) the priority of geometric design parameters for four types of combined curves were different. Based on the results, drivers and autonomous vehicles can pay different levels of attention to their speed change and departure behavior on different combination curves, and take targeted improvement measures in time according to the driving status of the vehicles. Road designers can also prioritize more important road design parameters in the design process to avoid serious accidents caused by excessive speed changes and departures.
In-vehicle infotainment systems can increase cognitive load and impair driving performance. These effects can be alleviated through interfaces that can assess cognitive load and adapt accordingly. Eye-tracking and physiological measures that are sensitive to cognitive load, such as pupil diameter, gaze dispersion, heart rate (HR), and galvanic skin response (GSR), can enable cognitive load estimation. The advancement in cost-effective and nonintrusive sensors in wearable devices provides an opportunity to enhance driver state detection by fusing eye-tracking and physiological measures. As a preliminary investigation of the added benefits of utilizing physiological data along with eye-tracking data in driver cognitive load detection, this paper explores the performance of several machine learning models in classifying three levels of cognitive load imposed on 33 drivers in a driving simulator study: no external load, lower difficulty 1-back task, and higher difficulty 2-back task. We built five machine learning models, including k-nearest neighbor, support vector machine, feedforward neural network, recurrent neural network, and random forest (RF) on (1) eye-tracking data only, (2) HR and GSR, (3) eye-tracking and HR, (4) eye-tracking and GSR, and (5) eye-tracking, HR, and GSR. Although physiological data provided 1%–15% lower classification accuracies compared with eye-tracking data, adding physiological data to eye-tracking data increased model accuracies, with an RF classifier achieving 97.8% accuracy. GSR led to a larger boost in accuracy (29.3%) over HR (17.9%), with the combination of the two factors boosting accuracy by 34.5%. Overall, utilizing both physiological and eye-tracking measures shows promise for driver state detection applications.
For automated driving at SAE level 3 or lower, driver performance in responding to takeover requests (TORs) is decisive in providing system safety. A driver state monitoring system that can predict a driver’s performance in a TOR event will facilitate a safer control transition from vehicle to driver. This experimental study investigated whether driver eye-movement measured before a TOR can predict driving performance in a subsequent TOR event. We recruited participants (N = 36) to obtain realistic results in a real-vehicle study. In the experiment, drivers rode in an automated vehicle on a test track for about 32 min, and a critical TOR event occurred at the end of the drive. Eye movements were measured by a camera-based driver monitoring system, and five measures were extracted from the last 2-min epoch prior to the TOR event. The correlations between each eye-movement measure and driver reaction time were examined, and a multiple regression model was built using a stepwise procedure. The results showed that longer reaction time could be significantly predicted by a smaller number of large saccades, a greater number of medium saccades, and lower saccadic velocity. The implications of these relationships are consistent with previous studies. The present real-vehicle study can provide insights to the automotive industry in the search for a safer and more flexible interface between the automated vehicle and the driver.
In China, F-type-5 m intersections are not uncommon. One approach of these intersections usually includes a driveway closely followed by an intersecting street, and the driveway and the intersecting street are parallel and approximately 5 m apart. Nowadays, drivers often rely on the navigation systems for directions. However, it is found that the navigation systems sometimes mislead or confuse drivers to make wrong turns or miss their turns at such F-type-5 m intersections. This study proposed to employ driving simulation to identify the appropriate prompt message delivered at the right prompt timing to help drivers navigate through such F-type-5 m intersections. First, a within-subjects two-factor experiment was designed. One factor was the Prompt Timing Mode (PTM), representing a set of three sequential messages broadcast by the navigation system at varying distances to the intended intersection; the other factor was the Prompt Message Type (PMT), representing various sets of three sequential messages broadcast by the navigation system. Three Prompt Timing Modes were used: PTM1 = {− 400 m, -200 m, − 30 m}, PTM2 = {− 300 m, − 150 m, − 30 m}, and PTM3 = {− 200 m, − 100 m, − 30 m}. Three Prompt Message Types were defined: PMT-A = {Turn right at the traffic light XXm ahead; Turn right at the traffic light XXm ahead; Turn right}, PMT-B = {Turn right at the traffic light XXm ahead, enter YY street; Turn right at the traffic light XXm ahead, enter YY street; Turn right}, PMT-C = {Turn right at the traffic light XXm ahead, enter YY street, and please use the second right turn lane; Turn right at the traffic light XXm ahead, enter YY street, and please use the second right turn lane; Turn right}. The combinations of the two factors generated nine experimental intersections which were randomly assigned to three experimental routes. Then, a total of 37 drivers were recruited, and participated in the driving simulation experiment from which vehicle operation data were collected under different prompt timing modes and message types. Next, the repeated Analysis of Variance (rANOVA) was performed to examine the effects of different prompt timing modes and prompt message types on vehicle operation indicators, such as Driving Time, Standard Deviation of Speed, Absolute Value of Acceleration, and Standard Deviation of Acceleration. Finally, the grey near-optimal method was adopted to evaluate the effectiveness of three prompt message types under each prompt timing mode. The rANOVA results showed the vehicle operation in the F-type-5 m intersection was affected by prompt timing modes and prompt message types; the evaluation results indicated that PMT-C made drivers perform better in PTM1 and PTM3, while PMT-B made drivers perform better inPTM2. However, the effectiveness of PMT-A was the lowest in each prompt timing mode. The research results provide valuable guidance to design the human machine interface of navigation systems, which can help drivers safely navigate through F-type-5 m intersections. This research also has laid solid foundations for establishing navigation messaging design guidelines.
This study aims to evaluate the risk level associated with the geometric parameters of two-lane horizontal curves. It is measured using critical average lateral acceleration (ac), a new performance index derived from lateral acceleration profile data and representing a normalized lateral acceleration beyond a threshold value. The required data for the study were obtained from a fixed-base driving simulator in which 41 drivers drove through the geometric configurations comprising 26 horizontal curves. The hierarchical clustering analysis provided three risk clusters for ac values represented as low-, moderate-, and high-risk events. These risk clusters were analyzed for the geometric parameters, such as radius, design speed, gradient, and preceding tangent length. The cross-tabulation results indicated that the curve radius less than 100 m represented 2-10 times higher crash risk than the curves with a larger radius (> 100 m). The curves on the descending gradient exhibited two times higher risk than the one on the flat and ascending gradient. Further, the decision tree provided design speed and its interaction with gradient and preceding tangent length as the significant parameters to assess risk along curves. Overall, this study establishes the suitability of the newly developed performance parameter as a surrogate safety measure for evaluating the risk associated with different geometric configurations.
Real-time crash risk prediction is expected to play a crucial role in preventing traffic accidents. However, most existing studies only focus on freeways rather than urban arterials. This paper proposes a real-time crash risk prediction model on arterials using a long short-term memory convolutional neural network (LSTM-CNN). This model can explicitly learn from the various features, such as traffic flow characteristics, signal timing, and weather conditions. Specifically, LSTM captures the long-term dependency while CNN extracts the time-invariant features. The synthetic minority over-sampling technique (SMOTE) is used for resampling the training dataset. Five common models are developed to compare the results with the proposed model, such as the XGBoost, Bayesian Logistics Regression, LSTM, etc. Experiments suggest that the proposed model outperforms others in terms of Area Under the Curve (AUC) value, sensitivity, and false alarm rate. The findings of this paper indicate the promising performance of using LSTM-CNN to predict real-time crash risk on arterials.
This study designs a framework of feature extraction and selection, to assess vehicle driving and predict risk levels. The framework integrates learning-based feature selection, unsupervised risk rating, and imbalanced data resampling. For each vehicle, about 1300 driving behaviour features are extracted from trajectory data, which produce in-depth and multi-view measures on behaviours. To estimate the risk potentials of vehicles in driving, unsupervised data labelling is proposed. Based on extracted risk indicator features, vehicles are clustered into various groups labelled with graded risk levels. Data under-sampling of the safe group is performed to reduce the risk-safe class imbalance. Afterwards, the linkages between behaviour features and corresponding risk levels are built using XGBoost, and key features are identified according to feature importance ranking and recursive elimination. The risk levels of vehicles in driving are predicted based on key features selected. As a case study, NGSIM trajectory data are used in which four risk levels are clustered by Fuzzy C-means, 64 key behaviour features are identified, and an overall accuracy of 89% is achieved for behaviour-based risk prediction. Findings show that this approach is effective and reliable to identify important features for driving assessment, and achieve an accurate prediction of risk levels.
This paper is aimed at obtaining a better understanding of driving behavior on horizontal curves of two-lane rural highways in terms of trajectories in relation to the different curve radii and directions by a driving simulator experiment. The driving simulator experiment involved 50 drivers and eight classes of curve radii, ranging from 125 m to 800 m. Overall, 2000 curve trajectories were analyzed and classified. Six major classes were defined: (1) ideal behavior, (2) normal behavior, (3) driving close to the centerline, (4) driving outside in curve approach, (5) cutting, and (6) correcting. Furthermore, 21 sub-classes were introduced to consider both lane departures and location of the corrective actions. The CATANOVA tests and Bhapkar’s tests showed that both the curve radius and the curve direction had a significant effect on the classification results. To get a clearer understanding of the effect of the curve radius and direction on curve negotiation, three macro-classes corresponding to safe, intermediate, and dangerous behavior were introduced. The safest behaviors significantly increased with the curve radius while the most dangerous behaviors significantly decreased with the curve radius. Furthermore, left curves showed a higher proportion of dangerous trajectories. Overall, it seems that the driving trajectories are a promising surrogate measure of safety as highlighted by the correlation between the trajectories identified as dangerous and the radii of the curves.
Recent advances in intelligent transportation system allow traffic safety studies to extend from historic data-based analyses to real-time applications. The study presents a new method to predict crash likelihood with traffic data collected by discrete loop detectors as well as the web-crawl weather data. Matched case–control method and support vector machines (SVMs) technique were employed to identify the risk status. The adaptive synthetic over-sampling technique was applied to solve the imbalanced dataset issues. Random forest technique was applied to select the contributing factors and avoid the over-fitting issues. The results indicate that the SVMs classifier could successfully classify 76.32% of the crashes on the test dataset and 87.52% of the crashes on the overall dataset, which were relatively satisfactory compared with the results of the previous studies. Compared with the SVMs classifier without the data, the SVMs classifier with the web-crawl weather data increased the crash prediction accuracy by 1.32% and decreased the false alarm rate by 1.72%, showing the potential the potential value of the massive web weather data. Mean impact value method was employed to evaluate the variable effects, and the results are identical with the results of most of previous studies. The emerging technique based on the discrete traffic data and web weather data proves to be more applicable on real-time safety management on freeways.
To improve highway design consistency, several studies developed operating speed prediction models and investigated drivers' speed behavior. Most existing models are based on spot speed data that assume constant operating speed throughout the horizontal curves and occurrence of acceleration and deceleration only on tangents. To overcome limitations associated with existing models, this study investigated continuous speed profiles with an experiment that used a high-fidelity dynamic driving simulator on a two-lane highway. A piecewise linear regression model and locally weighted regression scatter-plot smoothing were used to remove noise in the data set while preserving underlying patterns and to identify significant changes in the speed profile. Based on the smoothed speed profiles, models to predict operating speed in curves and in tangents, deceleration and acceleration rates to be used in the operating speed profiles, and starting and ending points of constant operating speed in curve were developed. Radius of the curve notably affected not only the operating speed in the curve but also the operating speed of the tangent following the curve: the smaller the radius, the lower the operating speed of the exit tangent. Both acceleration and deceleration rates increased with curvature. This study found that operating speed was not constant along curves. On small radius curves, deceleration ended close to the center of the curve, and acceleration starts, close to the end of the curve. Increasing the curve radius, the end point of deceleration moves toward the curve's beginning, whereas the start of acceleration moves toward the center of the curve.
Learning to store information over extended time intervals by recurrent backpropagation takes a very long time, mostly because of insufficient, decaying error backflow. We briefly review Hochreiter's (1991) analysis of this problem, then address it by introducing a novel, efficient, gradient based method called long short-term memory (LSTM). Truncating the gradient where this does not do harm, LSTM can learn to bridge minimal time lags in excess of 1000 discrete-time steps by enforcing constant error flow through constant error carousels within special units. Multiplicative gate units learn to open and close access to the constant error flow. LSTM is local in space and time; its computational complexity per time step and weight is O. 1. Our experiments with artificial data involve local, distributed, real-valued, and noisy pattern representations. In comparisons with real-time recurrent learning, back propagation through time, recurrent cascade correlation, Elman nets, and neural sequence chunking, LSTM leads to many more successful runs, and learns much faster. LSTM also solves complex, artificial long-time-lag tasks that have never been solved by previous recurrent network algorithms.
Driving style may have an important effect on traffic safety. Proactive crash risk prediction for lane-changing behaviors incorporating individual driving styles can help drivers make safe lane-changing decisions. However, the interaction between driving styles and lane-changing risk is still not fully understood, making it difficult for advanced driver-assistance systems (ADASs) to provide personalized lane-changing risk information services. This paper proposes a personalized risk lane-changing prediction framework that considers driving style. Several driving volatility indices based on vehicle interactive features have been proposed, and a dynamic clustering method is developed to determine the best identification time window and methods of driving style. The Light Gradient Boosting Machine (LightGBM) based on Shapley additive explanation is used to predict lane-changing risk for cautious, normal, and aggressive drivers and to analyze their risk factors. The highD trajectory dataset is used to evaluate the proposed framework. The obtained results show that i) spectral clustering and a time window of 3 s can accurately identify driving styles during the lane-changing intention process; ii) the LightGBM algorithm outperforms other machine learning methods in personalized lane-changing risk prediction; iii) aggressive drivers seek more individual driving freedom than cautious and normal drivers and tend to ignore the state of the car behind them in the target lane, with a greater lane-changing risk. The research conclusion can provide basic support for the development and application of personalized lane-changing warning systems in ADASs.
Illegal running into the opposite lane (IROL) on curve sections of two-lane rural roads is a frequently hazardous behavior and highly prone to fatal crashes. Although driving behaviors are always determined by the information from drivers' visual perceptions, current studies do not consider visual perceptions in predicting the occurrence of IROL. In addition, most machine learning methods belong to black-box algorithms and lack the interpretation of prediction results. Therefore, this study aims to propose an interpretable prediction model of IROL on curve sections of two-lane rural roads from drivers' visual perceptions. A new visual road environment model, consisting of five different visual layers, was established to better quantify drivers' visual perceptions by using deep neural networks. In this study, naturalistic driving data was collected on curve sections of typical two-lane rural roads in Tibet, China. There were 25 input variables extracted from the visual road environment, vehicle kinematics, and driver characteristics. Then, XGBoost (eXtreme Gradient Boosting) and SHAP (SHapley Additive exPlanation) methods were combined to build a prediction model. The results showed that our prediction model performed well, with an accuracy of 86.2% and an AUC value of 0.921. The average lead time of this prediction model was 4.4 s, sufficient for drivers to respond. Due to the advantages of SHAP, this study interpreted the impacting factors on this illegal behavior from three aspects, including relative importance, specific impacts, and variable dependency. After offering more quantitative information on the visual road environment, the findings of this study could improve the current prediction model and optimize road environment design, thereby reducing IROL on curve sections of two-lane rural roads.
Traffic crashes typically occur in a few seconds and real-time prediction can significantly benefit traffic safety management and the development of safety countermeasures. This paper presents a novel deep learning model for crash identification based on high-frequency, high-resolution continuous driving data. The method consists of feature engineering based on Convolutional Neural Network (CNN) and Gated Recurrent Unit (GRU) and classification based on Extreme Gradient Boosting (XGBoost). The CNN-GRU architecture captures the time series characteristics of driving kinematics data. Compared to normal driving segments, safety-critical events (SCEs)—i.e., crashes and near-crashes (CNC)—are rare. The weighted categorical cross-entropy loss and oversampling methods are utilized to address this imbalance issue. An XGBoost classifier is utilized instead of the multi-layer perceptron (MLP) to achieve a high precision and recall rate. The proposed approach is applied to the Second Strategic Highway Research Program Naturalistic Driving Study (SHRP 2 NDS) data with 1,820 crashes, 6,848 near-crashes, and 59,997 normal driving segments. The results show that in a 3-class classification system (crash, near-crash, normal driving segments), the accuracy for the overall model is 97.5%, and the precision and recall for crashes are 84.7%, and 71.3% respectively, which is substantially better than benchmarks models. Furthermore, the recall of the most severe crashes is 98.0%. The proposed crash identification approach provides an accurate, highly efficient, and scalable way to identify crashes based on high frequency, high-resolution continuous driving data and has broad application prospects in traffic safety applications.
The real-time crash potential prediction model is one of the important components of proactive traffic management systems. Over the years numerous models have been proposed to predict crash potential and achieved promising results using input data from roadside detectors. However, the detectors are normally installed at certain locations with limited coverage, while the connected vehicle data can provide city-wide mobility information. Previous studies have found that driver event variables such as hard braking, hard accelerations, etc. are correlated with crash potential on the road segments. Nevertheless, the existing studies are mostly conducted at the aggregated level, and the data are mostly collected from commercial vehicles such as taxis or buses traveling in the urban areas. This paper proposes a bidirectional long short-term memory (LSTM) model with two convolutional layers to predict real-time crash potential on freeways. The input data including traffic flow variables from detectors, and driver event variables from connected vehicle (CV) data, are aggregated at the one-minute level. The model achieves a recall value of 0.772 and an AUC value of 0.857. Moreover, to investigate the transferability of the proposed model, the original data are aggregated at the hourly level. The transferred model is developed with fine tuning two convolutional layers of the established model. And the transferred model achieves a recall value of 0.715 and an AUC value of 0.763. This proves that the proposed model can be successfully applied to another similar data set, or when the connected vehicles have lower penetration rate. In this study, we proved the usefulness of the connected vehicle data in the prediction of real-time crash potential, and the possibility of using it without detector data once the penetration rate increases to a reasonable level.
Horizontal curves are typically associated with increased crash risk when compared with straight roads, but recent analyses have suggested that having more frequent sharp curves decreases the relative crash risk posed by each curve. Here, 90 drivers completed a simulated rural drive with either high proximity (160 m straight tangent between curves) or low proximity (1200 m tangent) curves. Curve proximity had a significant effect on approach speeds, with drivers in the high proximity curve drive showing significantly lower mean and maximum approach speeds before entering the curve. However, they also showed an unexpected tendency to higher speeds while negotiating the curve itself. The current study provides direct empirical evidence that driving behaviour on approach to a given curve is significantly affected by the proximity of other curves, and therefore highlights the need to factor in the characteristics of the road on approach to the curve, as well as the features of the curve itself when assessing risk.
An inferred design speed (IS) is the maximum vehicle speed for which all critical design speed-related criteria are met at a particular location, which is widely recognized as an important speed limit on highways. However, a side friction factor is used as an index for vehicle lateral stability in the calculation of IS for horizontal curves. This condition causes inaccuracies in the value of IS due to the incapability of the index in fully characterizing the steering stability and track-holding capability of vehicles. This paper aims to fill this gap by discovering the lateral instability modes of vehicles on horizontal curves and to determine the boundary of each index beyond which the corresponding mode of instability occurs. Thus, a theoretical framework for realizing a revised IS (RIS) was established. A high-precision seven-degree-of-freedom driver–vehicle–road mathematical model was developed using MATLAB/Simulink. The state of vehicle lateral instability on highway horizontal curves was decoupled, and the boundaries of vehicle stability indices were deduced using the theory of vehicle system dynamics. Accordingly, a procedure for determining RIS was proposed. Case studies were conducted on the three horizontal curves of a two-lane rural highway in China, and existing speed limit methods were compared with RIS. With in-depth consideration of the lateral instability characteristics of vehicles, RIS provides traffic management departments with a precise maximum safe speed on horizontal curves under various pavement friction conditions. It is expected to be a restriction with respect to safety for other speed limit strategies that consider driver expectancy and efficiency.
Lane change identification based on long short-term memory (LSTM) neural networks has received increasing attention. The most commonly known disadvantage of the approach is the “black box” nature. This study, to improve the interpretability, proposed a novel method for identifying lane changes. The philosophy was to measure the similarity of the glance behavior during the maneuver. The glance behavior in the same maneuver is more similar, while that in different maneuvers is different. First, a driving simulator was used to collect driving behavior data. The eye gaze time series were captured by the eye tracker. The median of the eye gaze sequence of 3 s before the initial moment of lane change was obtained. With the median as the center, the driver's visual field plane was divided into several grids with the side length of 600 pixels. A sliding space-time cuboid algorithm was proposed to extract the scanning path. Second, four-dimensional dynamic time warping (4DDTW) distance was used to measure the similarity of the scanning paths. Third, the K nearest neighbor (KNN) algorithm was used to classify the driving maneuver into the lane-keeping (LK), the right lane change (RLC), and the left lane change (LLC) based on the 4DDTW distance of the scan paths. LSTM was also used to classify the driving maneuver into LK, RLC, and LLC. The performances of 4DDTW-KNN and LSTM were compared. The accuracies of 4DDTW-KNN and LSTM were 86.50% and 86.33%, respectively. LSTM is not better than 4DDTW-KNN in lane change identification based on eye gaze data with a comprehensive consideration of the time efficiency, the accuracy and the interpretability.
The drastic changes of the space environment at the tunnel entrance can lead to frequent accidents with higher levels. The connected vehicle environment provides drivers with surrounding traffic information and improve their driving behavior by helping them make safe decisions efficiently. As such, this study is to examine the effects of the connected vehicle environment on driving behavior and safety at the tunnel entrance zone. To this end, this research simulates a connected vehicle environment and provides driving aids through the Human-Machine Interface (HMI). Secondly, 40 participants with diverse backgrounds drove the simulator under two different driving conditions: HMI-OFF (traditional driving environment) and HMI-ON (connected vehicle environment). Finally, indicators are selected from speed control, stability and urgency to analyze the impact of the connected vehicle environment on drivers’ behaviors and safety at the warning zone and tunnel entrance zone. The results show that in the connected vehicle environment, the drivers’ speed control in the warning zone is improved and their deceleration behavior is advanced. The driver's speed control and stability are improved while the danger level of the accident is reduced 100 m ahead of the tunnel entrance. Besides, the driver's speed control and stability have been both improved 300 m after the tunnel entrance. Overall, in the connected vehicle environment, the driver can recognize the tunnel in advance and adjust his driving speed in time to ensure his safety at the tunnel entrance. The results of this study play a critical role in the design and research of warning systems in a connected vehicle environment, and will also guide vehicle manufacturers in designing safety-related functions of automated vehicles. In this research, a connected vehicle environment test platform based on driving simulation technology is constructed and tested in specific tunnel entrance scenarios, which provides a reference for realizing active protection of vehicles at the tunnel entrance.
Real-time driving risk status prediction is critical for developing proactive traffic intervention strategies and enhance driving safety. However, the optimal observation time window length and prediction time window length, which should be the prerequisite for the timeliness and accuracy of real-time driving risk status prediction model, have been rarely explored in previous studies. In this study, a methodology which integrates driving risk status identification, rolling time window-based feature extraction, real-time driving risk status prediction and driving risk influencing factors analysis was proposed to accurately evaluate and predict real-time driving risk status. The methodology was tested based on 1,440 car-following events from Shanghai Naturalistic Driving Study. Results show that four driving risk statuses (safe, low-risk, median-risk and high-risk) are most appropriate to establish risk labelling criteria. In addition, results from driving risk status prediction show that when the observation time window length is 0.5 s, the accuracy rate of predicting medium-risk or high-risk status occurring in the next 0.7 s is higher than 85 % using multi-layer perceptron model. Meanwhile, the results from the analysis of influencing factors show that the input variables related to the risk status score higher in the ranking of feature importance. A part from that, speed difference, headway distance, speed and acceleration are still important in predicting driving risk status. The proposed methods in this paper can be applied in connected and autonomous vehicle (CAV) to reduce driver cognitive workload and hence improve driving safety fed with naturalistic driving data collected using in-vehicle systems.
Car manufacturers expect driving simulators to be reliable research and development tools. Questions arise, however, as to whether drivers’ behavior on simulators exactly matches that observed when they are driving real cars. Drivers’ performances and their subjective feelings about their driving were compared between two groups during a 40-min driving test on the same circuit in a real car (n = 20) and a high-fidelity dynamic simulator (n = 27). Their speed and its variability, the braking force and the engine revolutions per minute (rpm) were recorded five times on a straight line and three times on a curve. The differences observed in these measurements between circuit driving (CD) and simulator driving (SD) from the 6th to 40th minute showed no significant changes during the drive. The drivers also completed the NASA Raw Task Load Index (NASA RTLX) questionnaire and the Simulator Sickness Questionnaire (SSQ) and estimated the ease and standard of their own driving performances. These subjective feelings differed significantly between the two groups throughout the experiment. The SD group’s scores on the NASA RTLX and SSQ questionnaires increased with time and the CD group’s perceived driving quality and ease increased with time, reaching non-significantly different levels from their usual car driving standards by the end of the drive. These findings show the existence of a fairly good match between real-life and simulated driving, which stabilized six minutes after the start of the test, regardless of whether the road was straight or curved. These objective findings and subjective assessments suggest possible ways of improving the match between drivers’ performances on simulators and their real-life driving behavior.
As a product of the shared economy, online car-hailing platforms can be used effectively to help maximize resources and alleviate traffic congestion. The driver’s behavior is characterized by his or her driving style and plays an important role in traffic safety. This paper proposes a novel framework to classify driving styles (defined as aggressive, normal, and cautious) based on online car-hailing data to investigate the distinct characteristics of drivers when performing various driving tasks (defined as cruising, ride requests, and drop-off) and undergoing certain maneuvers (defined as turning, acceleration, and deceleration). The proposed model is constructed based on the detection and classification of driving maneuvers using a threshold-based endpoint detection approach, principal component analysis, and k-means clustering. The driving styles that the driver exhibits for the different driving tasks are compared and analyzed based on the classified maneuvers. The empirical results for Nanjing, China demonstrate that the proposed framework can detect driving maneuvers and classify driving styles accurately. Moreover, according to this framework, driving tasks lead to variations in driving style, and the variations in driving style during the different driving tasks differ significantly for turning, acceleration, and deceleration maneuvers.
Statistics show that horizontal curves, especially those of radii less than 200 m, present an increased road accident risk mainly due to inappropriate speed and failure to maintain proper lateral position. This simulator study aims to analyse how two low-cost road marking measures (red median and horizontal warning signs), alone or combined with a vertical warning sign, affect driver behaviour (driving speed, lateral movement, acceleration/deceleration) before and throughout dangerous horizontal curves on a two-way rural road. With GIS-supported mapping of traffic accidents, we identified the most dangerous curves on the main rural road in Croatia and replicated them on the driving simulator. Based on the driving runs of 43 participants, the study concluded that both measures, used either alone or combined with a vertical warning sign, significantly reduced the speed compared to the control condition (vertical warning sign alone). Additionally, the use of a red median prompted the lateral movement of the vehicle closer to the edge line. The paper also defines the potential use of the measures for dealing with specific types of curve-related accidents.
Background
Many studies have found that eye movement behavior provides a real-time index of mental activity. Risk management architectures embedded in autonomous vehicles fail to include human cognitive aspects. We set out to evaluate whether eye movements during a risk driving detection task are able to predict risk situations.
Methods
Thirty-two normally sighted subjects (15 female) saw 20 clips of recorded driving scenes while their gaze was tracked. They reported when they considered the car should brake, anticipating any hazard. We applied both a mixed-effect logistic regression model and feedforward neural networks between hazard reports and eye movement descriptors.
Results
All subjects reported at least one major collision hazard in each video (average 3.5 reports). We found that hazard situations were predicted by larger saccades, more and longer fixations, fewer blinks, and a smaller gaze dispersion in both horizontal and vertical dimensions. Performance between models incorporating a different combination of descriptors was compared running a test equality of receiver operating characteristic areas. Feedforward neural networks outperformed logistic regressions in accuracies. The model including saccadic magnitude, fixation duration, dispersion in ×, and pupil returned the highest ROC area (0.73).
Conclusion
We evaluated each eye movement descriptor successfully and created separate models that predicted hazard events with an average efficacy of 70% using both logistic regressions and feedforward neural networks. The use of driving simulators and hazard detection videos can be considered a reliable methodology to study risk prediction.
Connected Vehicles (CV) technology has been used to address safety issues on highway horizontal curves. Existing curve warning systems are either using curve warning signs or providing drivers with an in-vehicle curve warning message in advance, allowing drivers to adjust their speed prior to the vehicle entering the curve. In practice, drivers might be compliant before entering the curve but may pick up the speed in the curve. Therefore, it remains a problem that existing curve warning systems are not able to guide drivers by providing necessary speed warnings through the entire course of approaching, entering, navigating, and leaving horizontal curves. Therefore, the objective of this study is to improve curve speed compliance by proposing a guidance-oriented Advanced Curve Speed Warning system (Advanced-CSW) with a focus on providing guided curve speed messages throughout the horizontal curves. The Advanced-CSW system is based on Dedicated Short-Range Communication (DSRC) enabling vehicle-infrastructure (V2I) communication. Anytime the vehicle is speeding, the guided message will be displayed until the vehicle's speed is within compliant range. Drivers who use the Advanced-CSW can receive multiple guided messages from the in-vehicle heads-up display through the entire course of navigating through horizontal curves. Thirty participants are recruited to perform the driving experiment on the simulator of driving through a series of horizontal curves under various geometric, roadway and traffic conditions. These conditions include different curve severity, illumination, and pavement wetness levels. The Advanced-CSW system's performance was evaluated in terms of the speed difference, which measures the gap between the in-curve mean speed and curve advisory speed. The results were compared with the performance of speed difference by driving with CSW or CSO through the entire curve. The experiment data was modeled using the mixed linear model with random effects, which includes the individual's driving behavior. In summary, when male drivers navigate through the horizontal curves under different curve speed warning systems , their speed compliance is significantly increased with continuous and guided messages provided in comparison with the speed compliance under the one-time curve speed warning message and the curve sign only. Female drivers improve their speed compliance in the curve by using curve signs only comparing to using one-time curve speed warning message or continuous guided curve speed warning messages. Also, male drivers' speed differences by using the guided system are significantly reduced by 6.53~7.68 mi/h compared to driving with curve signs only or one-time curve speed warning message. In addition, there is also a speed reduction of 1.81 mi/h if male drivers receiving continuous guided messages in the curve during the daytime than during the nighttime. The proposed adaptive system based on that is adaptive to the vehicle's real-time speed and location by providing a new direction in designing effective curve warning systems. The speed-guided messages through the entire course of approaching, entering, navigating, and leaving horizontal curves can solve the current issue of speed incompliance by using the existing curve warning systems.
Real-time traffic crash prediction has been a major concern in the development of Collision Avoidance Systems (CASs) along with other intelligent and resilient transportation technologies. There has been a pronounced progress in the use of machine learning models for crash events assessment by the transportation safety research community in recent years. However, little attention has been paid so far to evaluating real-time crash occurrences within information fusion systems. The main aim of this paper is to design and validate an ensemble fusion framework founded on the use of various base classifiers that operate on fused features and a Meta classifier that learns from base classifiers’ results to acquire more performant crash predictions. A data-driven approach was adopted to investigate the potential of fusing four real-time and continuous categories of features namely physiological signals, driver maneuvering inputs, vehicle kinematics and weather covariates in order to systematically identify the crash strongest precursors through feature selection techniques. Moreover, a resampling-based scheme, including Bagging and Boosting, is conducted to generate diversity in learner combinations comprising Bayesian Learners (BL), k-Nearest Neighbors (kNN), Support Vector Machine (SVM) and Multilayer Perceptron (MLP). To ensure that the proposed framework provide powerful and stable decisions, an imbalance-learning strategy was adopted using the Synthetic Minority Oversampling TEchnique (SMOTE) to address the class imbalance problem as crash events usually occur in rare instances. The findings show that Boosting depicted the highest performance within the fusion scheme and can accomplish a maximum of 93.66% F1 score and 94.81% G-mean with Naïve Bayes, Bayesian Networks, k-NN and SVM with MLP as the Meta-classifier. To the best of our knowledge, this work presents the first attempt at establishing a fusing framework on the basis of data from the four aforementioned categories and fusion models while accounting for class imbalance. Overall, the method and findings provide new insights into crash prediction and can be harnessed as a promising tool to improve intervention efforts related to traffic intelligent transportation systems.
Speeding is a major traffic violation and time pressure is one of the leading contributors to speeding. High-speed driving requires an immediate response to perilous events from the driver to avoid a crash. Reaction time is one of the important driving performance measures to assess the driver’s response to the event. Therefore, the current study examined the influence of time pressure on reaction times of the drivers measured for two different perilous events (pedestrians crossing and obstacle overtaking). Eighty-five Indian licensed drivers participated in a driving simulation study designed for three different time pressure conditions: No Time Pressure (NTP), Low Time Pressure (LTP), and High Time Pressure (HTP). The survival analysis technique was used to model the effect of time pressure and driver characteristics with reaction times of the drivers. It was observed that drivers’ reaction times decreased by 18% and 9% in LTP and 28% and 16% in HTP during the pedestrians crossing and obstacle overtaking events, respectively. Further, 1 m/second increase in approach speed resulted in 2% and 4% reduction in reaction times of the drivers in pedestrians crossing and obstacle overtaking events, respectively. Young drivers responded 21% faster than mature drivers during the pedestrians crossing event. Interestingly, sleeping hours and physical fitness played an important role in driver’s reaction to the events. The drivers performing regular physical exercise and having minimum eight-hours of overnight sleep reacted 16% and 17% earlier in pedestrians crossing and obstacle overtaking events, respectively. The overall findings from this study showed enhanced stimulus-response behaviour of the drivers under time pressure driving conditions. The results obtained from the study can give new insight into various safety-related ITS applications.
Aggressive driving, amongst all driving behaviors, is largely responsible for leading to traffic accidents. With the objective to improve road safety, this paper develops an on-line approach for vehicle running state monitoring and aggressive driving identification, using kinematic parameters captured by the in-vehicle recorder under naturalistic driving conditions. To characterize the roads in reality, a novel road conceptual model is proposed. It accounts for not only the curve on the horizontal plane but also the slope on the vertical plane, as well as the cross slope. For each position where the vehicle is driving, the vehicle motion is decomposed into two circular motions on the horizontal and vertical planes. On each plane, the vehicle maneuver is first identified. Then, aggressive driving is identified according to the limit equilibrium of driving safety or comfortability. Based on the proposed method called “three-elements”, the vehicle maneuver, radius and slope angle on the vertical plane can be solved in an on-line manner. The novel approach is an elaborate analytical model with clear physical meaning but small computation load, and therefore is potential to be implemented in the mobile devices to assist in real-time aggressive driving identification and labeling. The developed approach is applied to a real case on the curved and sloped route in Nanjing, China. Empirical results of extensive experiments, based on the kinematic parameters collected from the in-vehicle data recorder under naturalistic driving conditions, demonstrate that aggressive driving behaviors are mostly found on the pavements with curve and slope, and can be identified by the developed approach.
Introduction: An improper driving strategy is one of the causative factors for a high probability of runoff and overturning crashes along the horizontal curves of two-lane highways. The socio-demographic and driving experience factors of a driver do influence driving strategy. Hence, this paper explored the effect of these factors on the driver’s runoff risk along the horizontal curves.
Method: The driving performance data of 48 drivers along 52 horizontal curves was recorded in a fixed-base driving simulator. The driving performance index was estimated from the weighted lateral acceleration profile of each driver along a horizontal curve. It was clustered and compared with the actual runoff events observed during the experiment. It yielded high, moderate and low-risk clusters. Using cross-tabulation, each risk cluster was compared with the socio-demographic and experience factors. Further, generalized mixed logistic regression models were developed to predict the high-risk and high to moderate risk events.
Results: The age and experience of drivers are the influencing factors for runoff crash. The high-risk event percentage for mid-age drivers decreases with an increase in driving experience. For younger drivers, it increases initially but decreases afterwards. The generalized mixed logistic regression models identified young drivers with mid and high experience and mid-age drivers with low-experience as the high-risk groups.
Conclusions: The proposed index parameter is effective in identifying the risk associated with horizontal curves. Driver training program focusing the horizontal curve negotiation skills and graduated driver licensing could help the high-risk groups.
Practical applications: The proposed index parameter can evaluate driving behavior at the horizontal curves. Driving behavior of high-risk groups could be considered in highway geometric design. The motor vehicle agencies, advanced driver assistance systems manufacturers and insurance agencies can use proposed index parameter to identify the high-risk drivers for their perusal.
This paper focuses on the behaviours adopted by road users when negotiating horizontal curves with sight limitations. Experiments at a driving simulator were conducted on two-lane highways in which drivers were confronted with a range of sight conditions generated by the manipulation of variables such as curve direction, radii and distance of lateral sight obstructions along horizontal curves. It was observed that most of the drivers adopted strategies which resulted in a stopping distance shorter than the available sight distance, thereby maintaining safe driving conditions. Some drivers reduced their speed, some increased the lateral distance from any sight obstructions along the roadside, some did both, while others did neither. A preliminary analysis indicated that the safety benefits resulting from a vehicle speed reduction strategy significantly outweigh those from a lateral shift in the lane. Further analyses on the 1246 cases investigated offered further support for this proposition, while revealing that a higher proportion of drivers opted for the first strategy for safety reasons. Moreover, visibility conditions (safe, partially safe, and unsafe) played a role in the choice of driving strategies. Results provide evidence that a significant group of drivers used the two strategies under severely restricted visibility conditions (i.e., along sharp radius curves); however, the strategies selected were independent of the driver speed profile (i.e., slower, average, or faster).
With the help of traffic detectors widely deployed along arterial roads and intersections, real-time traffic data are collected and updated in a very short time period, which makes it possible to conduct real-time analysis at signalized intersections. Among them, real-time crash risk prediction is one of the most promising and challenging research topics. This study attempts to predict real-time crash risk by considering time series dependency with the employment of a long short-term memory recurrent neural network (LSTM-RNN) algorithm. Also, the synthetic minority over-sampling technique (SMOTE) was utilized in this study to generate a balanced training dataset for algorithm training. In comparison, a conditional logistic model was developed based on matched case control design. Both models were evaluated based on the real-world unbalanced test dataset rather than an artificially balanced dataset. The comparison results indicate that the LSTM-RNN with SMOTE outperforms the conditional logistic model. The methods and findings of this study attempt to verify the feasibility of real-time crash risk prediction by using LSTM-RNN with over-sampled dataset (SMOTE).
Combined horizontal and crest vertical curves are among the most hazardous road segments because of drivers’ difficulties in perceiving early road geometry. This study evaluates new treatments to improve the safety of horizontal and crest vertical curves and compares their efficiency for driver performance based on design consistency criteria under free flow traffic and on-coming traffic. A combination of chevron signs with three promising treatments included herringbones, sealed shoulder and a yellow blinking signal was applied to hazardous curve sections in a driving simulator. Performance measures consisted of mean of speed and lateral position. The results indicated that combining chevrons and a warning blinking signal is the most appropriate treatment for horizontal and crest vertical curves, as this resulted in lower speed and lateral position. In addition, speed and lateral position variations along the curves were lower compared to other treatments specially, with an on-coming vehicle. Sealed shoulder makes drivers drive faster while entering a curve and brake suddenly while changing direction. Using herring bones is found to reduce speed and speed differential along the curve but not lateral position which was even higher in treated curves particularly, in presence of an on-coming vehicle.
Mobile phone distraction has been recognized as an adverse factor that degrades drivers’ performance on road. Although research showed that drivers take various compensatory strategies to minimize the risk in distracted driving, little consensus has been achieved regarding the actual change in collision risk because of compensatory behaviours. This study aims to investigate the impact of mobile phone use and drivers’ compensatory behaviours on the collision risk in a car-following situation. By using a high-fidelity driving simulator, 37 participants completed the simulation experiment in three mobile phone use conditions: no phone (baseline), hands-free and hand-held. Cluster analysis was adopted to classify the final collision risk into different levels. Two logit regression models were developed to examine the relationships between drivers’ characteristics, mobile phone use, collision avoidance performances and their involvement in the collision risk. Results show that compared to no phone and hands-free, drivers using hand-held phone had a longer brake reaction time and also an increased likelihood of being involved in a high risk group. Drivers compensated to reduce the likelihood of safety-critical events through a simultaneous control of car-following speed and distance (i.e. Time-to-collision (TTC)) in distracted condition. Additionally, the results also indicated that female drivers and non-professional drivers were more likely to be involved in high risk group than male drivers and professional drivers. The study provided a systematic method to quantify the impact of mobile phone distraction and drivers’ compensation behaviors on collision risk. The effectiveness of compensatory strategy by controlling TTC also shed light on the development of intelligent transport systems to help distracted drivers avoid safety-critical situations.
Inappropriate speed selection on a curved road is a main cause of rollover accidents for heavy vehicles due to their relative higher centers of gravity, comparing with those of passenger cars. Traditional driving safety improvement methods on curves include static/dynamic roadside speed limit signs that lack individual vehicle's characteristics, and the high-cost anti-rollover stability control systems that cannot take road geometric parameters like superelevation of a vehicle's upcoming curve into consideration. In this paper,a new rollover speed prediction model based on the derivation of three-degree-of-freedom vehicle dynamics and lateral load transfer ratio (LTR) index is presented. Through numerical experiments, the results show that this model could guarantee the vehicle roll stability with the calculated speed for entering a curve whose road radius is even 50 m, in which the vehicle's LTR never exceeds 0.72 and lateral acceleration is always less than 0.63 g. Moreover, the proposed model built in a mobile smartphone app can calculate curve radius at first, then provide an early alarming to the driver with an appropriate speed if rollover accident is imminent on the curve. The field tests on freeway off-ramps show that this smartphone-based rollover speed warning system can calculate the curve radii, and alert the driver with appropriate curve speeds that are partially equivalent to professional skilled drivers’ speed choices.
Inappropriate curve speed influenced by the interactions of driver behaviours, vehicle dynamics and road environments is the dominant cause of vehicle lateral instability induced crashes, like sideslips and rollovers. The present study introduced a driver behaviour influence factor associated with drivers' driving styles comparing to a theoretical curve speed model that only considers the vehicle-road interaction. This factor is defined as the ratio of drivers' actual selected speed to the theoretical curve speed. Aiming at deriving the factor for different driving styles, it was utilized the 28-item Chinese version of Driver Behaviour Questionnaire (DBQ). A correlation analysis between DBQ subscales and the factor indicated that a driver with higher violations scores is prone to drive faster in curve negotiation. Based on this finding, 24 experienced professional drivers were classified into two types, i.e. the moderate and the aggressive, corresponding to their scores on DBQ violations scale. Through a simulation, it showed that the improved curve speed model could not only prevent the risks of rollover and sideslip, but also provided different appropriate curve safe speeds in accordance with drivers' driving styles.
Real-time crash prediction is the key component of the Vehicle Collision Avoidance System (VCAS) and other driver assistance systems. The further improvements of predictability requires the systemic estimation of crash risks in the driver-vehicle-environment loop. Therefore, this study designed and validated a prediction method based on the supervised learning model with added behavioral and physiological features. The data samples were extracted from 130 drivers’ simulator driving, and included various features generated from synchronized recording of vehicle dynamics, distance metrics, driving behaviors, fixations and physiological measures. In order to identify the optimal configuration of proposed method, the Discriminant Analysis (DA) with different features and models (i.e. linear or quadratic) was tested to classify the crash samples and non-crash samples. The results demonstrated the significant improvements of accuracy and specificity with added visual and physiological features. The different models also showed significant effects on the characteristics of sensitivity and specificity. These results supported the effectiveness of crash prediction by quantifying drivers’ risky states as inputs. More importantly, such an approach also provides opportunities to integrate the driver state monitoring into other vehicle-mounted systems at the software level.
To obtain the topological relation between vehicle travelling track and the highway geometric alignment, real vehicle on-road tests were carried out on seven mountainous two-lane highways; vehicle operation parameters under natural driving state were collected and travelling tracks were extracted. Continuous change curves of lateral offset rate of track (LORT) were obtained. Then the morphological features of vehicle tracks as well as the influence rules subjecting to road geometrical parameters were analyzed. The results show that when a car enters a left curve, it firstly veers to the inside of the curve, then moves towards the outside, and then drives back to its original lane. The track on right curve has a similar changing trends. However the positions of curve cutting and the positions car occupying the opposite lane are different; for curves with small or medium deflection angle, position of curve cutting of actual track appears earlier than the one of ideal track, and actual track is more close to the outside of curve; the longer the time of handling curves, the more likely the driver occupies the opposite lane, and this can lead to a greater probability of the secondary curve cutting; the car experiences a cutting may be too close to the outside of the curve because of its inertia, and its track needs to be corrected, however overcorrection may lead to the phenomenon that the car invades the opposite lane again; the smaller the curve radius, the wider encroachment into the opposite lane, and a left curve with a medium deflection angle has the greatest LORT, while for right curves LORT increases with the increase of deflection angle. Serious encroachment on opposite lane appears in switch-back curve and the largest degree of encroachment occurs in the right turn. © 2016, Editorial Department of China Journal of Highway and Transport. All right reserved.
The influence of driver distraction on driving performance is not yet well understood, but it can have detrimental effects on road safety. In this study, we examined the effects of visual and non-visual distractions during driving, using a high-fidelity driving simulator. The visual task was presented either at an offset angle on an in-vehicle screen, or on the back of a moving lead vehicle. Similar to results from previous studies in this area, non-visual (cognitive) distraction resulted in improved lane keeping performance and increased gaze concentration towards the centre of the road, compared to baseline driving, and further examination of the steering control metrics indicated an increase in steering wheel reversal rates, steering wheel acceleration, and steering entropy. We show, for the first time, that when the visual task is presented centrally, drivers’ lane deviation reduces (similar to non-visual distraction), whilst measures of steering control, overall, indicated more steering activity, compared to baseline. When using a visual task that required the diversion of gaze to an in-vehicle display, but without a manual element, lane keeping performance was similar to baseline driving. Steering wheel reversal rates were found to adequately tease apart the effects of non-visual distraction (increase of 0.5 degree reversals) and visual distraction with offset gaze direction (increase of 2.5 degree reversals). These findings are discussed in terms of steering control during different types of in-vehicle distraction, and the possible role of manual interference by distracting secondary tasks.
The coefficient of transverse force is an important index in evaluating driving safety and comfort. This paper takes the safety and comfort of curve sections in mountainous highways as a research object. As evaluation indexes, heart rate and systolic blood pressure are measured in many driving experiments, and their quantitative relationships with coefficient of transverse force are also analyzed and established. The reasonable value of the coefficient of transverse force should be less than 0.2.
The brake reaction time of drivers under real traffic risk scenarios is studied in this paper. Real traffic scenarios collected in Shanghai are screened and classified, and six typical types of risk scenarios are obtained. Driver brake reaction times under these six types of risk scenarios, especially the scenario of “front vehicle decelerating”, are measured and analyzed. The results show that driver brake reaction time is obviously different for different type of risk scenarios. Compared with the risk caused by vehicles, the driver brake reaction time for the risks caused by vulnerable road users is shorter. In the scenario of “front vehicle decelerating”, when the vehicle speed goes up, the risk urgency increases or the scene is not on the road intersection, the driver brake reaction time tends to reduce.
Horizontal curves are associated with significant numbers of fatal crashes. The safety performance of an individual curve depends on a variety of geometric factors, including radius, sup erelevation rate, deflection angle; vehicle speeds; and friction characteristics of the pavement. In addition, vehicle speeds in curves are influenced by curve geometry and speeds along the approach tangent. Hence, an understanding of the relationship between these various factors is important for developing an analysis framework to assess curve safety. hi this paper, documented models estimate vehicle speeds at the beginning, midpoint, and end of a curve on the basis of the curve's operational and geometric characteristics. These models are applied in an analysis framework based on the concept of margin of safety, which is side friction demand subtracted from side friction supply. This method can be used to prioritize proposed treatments at curve sites and to identify points along the curve at which sliding failures are most likely to occur. The analyst must provide the key operational and geometric characteristics as well as friction characteristics of the existing pavement and proposed surface treatment.
In order to improve the vehicle driving safety in curved roads, aimed at vehicle lateral unstable incidents in curves, such as rollover, sideslip, it was assumed that the information between vehicle and road could be interacted with cooperative vehicle infrastructure. Driver characteristics and vehicle structural parameters were introduced to a traditional model of safety speed calculation based on road environment. An improved model of safety speed calculation in curves was established, considering the influence of human, vehicle, and road. The experimental comparative analysis was conducted by using a certain type of truck in various conditions. The results show that safety speed of improved model always falls in between the calculated values of existing models. When road adhesion coefficient rises to a certain range, safety speed of improved model has a unique saturation phenomenon, which can reflect rollover accident occurred at high speed in curves.
In this paper, a road-departure warning unit taking into account driver-vehicle-infrastructure (DVI) interactions is proposed. The longitudinal and lateral vehicle dynamics limits are analyzed to detect the road departure on loss of control. Vehicle positioning and time to lane crossing (TLC) are used to detect the road departure on a defect of guidance. Prevention of excessive longitudinal speed is handled through the computation of a critical longitudinal speed when approaching a curve and speed profile generation in the straight road section preceding the curve. For the lateral mode, the vehicle oversteering or understeering, the yaw motion, and the lateral acceleration are analyzed. The vehicle lateral displacement and the TLC values are also examined when the vehicle dynamics are not excessive. Necessary data for detection algorithms, which are not available from measurements, are estimated using an extended Kalman filter. The system consists of several subsystems, which work in parallel and provide warning through a dedicated human-machine interface (HMI). This road-departure warning system is experimentally tested on a test track using a prototype vehicle. It is found to be efficient and robust.
Detecting Distraction Behavior of Drivers Using Naturalistic Driving Data. China
- Sun
Detecting Distraction Behavior of Drivers Using Naturalistic Driving Data
- J Sun
- Y Zhang
- J Wang
Sun, J., Zhang, Y., Wang, J., 2020. Detecting Distraction Behavior of Drivers Using
Naturalistic Driving Data. China. J Highw Transp 33, 225-235. https://doi.org/
10.19721/j.cnki.1001-7372.2020.09.022.