Feature selection for driving style and skill clustering using naturalistic driving data and driving behavior questionnaire.

Driver's driving style and driving skill have an essential influence on traffic safety, capacity, and efficiency. Through clustering algorithms, extensive studies explore the risk assessment, classification, and recognition of driving style and driving skill. This paper proposes a feature selection method for driving style and skill clustering. We create a supervised machine learning model of driver identification for driving behavior data with no ground truth labels on driving style and driving skill. The key features are selected based on permutation importance with the underlying assumption that the key features for clustering should also play an important role in characterizing individual drivers. The proposed method is tested on naturalistic driving data. We introduce 18 feature extraction methods and generate 72 feature candidates. We find five key features: longitudinal acceleration, frequency centroid of longitudinal acceleration, shape factor of lateral acceleration, root mean square of lateral acceleration, and standard deviation of speed. With the key features, drivers are clustered into three groups: novice, experienced cautious, and experienced reckless drivers. The ability of each feature to describe individuals' driving style and skill is evaluated using the Driving Behavior Questionnaire (DBQ). For each group, the driver's response to DBQ key questions and their distribution of key features are analyzed to prove the validity of the feature selection result. The feature selection method has the potential to understand driver's characteristics better and improve the accuracy of driving behavior modeling.

Risky Driving Behavior Recognition Based on Vehicle Trajectory.

This paper proposes a measurement of risk (MOR) method to recognize risky driving behavior based on the trajectory data extracted from surveillance videos. Three types of risky driving behavior are studied in this paper, i.e., speed-unstable driving, serpentine driving, and risky car-following driving. The risky driving behavior recognition model contains an MOR-based risk evaluation model and an MOR threshold selection method. An MOR-based risk evaluation model is established for three types of risky driving behavior based on driving features to quantify collision risk. Then, we propose two methods, i.e., the distribution-based method and the boxplot-based method, to determine the threshold value of the MOR to recognize risky driving behavior. Finally, the trajectory data extracted from UAV videos are used to validate the proposed model. The impact of vehicle types is also taken into consideration in the model. The results show that there are significant differences between threshold values for cars and heavy trucks when performing speed-unstable driving and risky car-following driving. In addition, the difference between the proportion of recognized risky driving behavior in the testing dataset compared with that in the training dataset is limited to less than 3.5%. The recognition accuracy of risky driving behavior with the boxplot- and distribution-based methods are, respectively, 91% and 86%, indicating the validation of the proposed model. The proposed model can be widely applied to risky driving behavior recognition in video-based surveillance systems.

Risky Driver Recognition with Class Imbalance Data and Automated Machine Learning Framework.

Identifying high-risk drivers before an accident happens is necessary for traffic accident control and prevention. Due to the class-imbalance nature of driving data, high-risk samples as the minority class are usually ill-treated by standard classification algorithms. Instead of applying preset sampling or cost-sensitive learning, this paper proposes a novel automated machine learning framework that simultaneously and automatically searches for the optimal sampling, cost-sensitive loss function, and probability calibration to handle class-imbalance problem in recognition of risky drivers. The hyperparameters that control sampling ratio and class weight, along with other hyperparameters, are optimized by Bayesian optimization. To demonstrate the performance of the proposed automated learning framework, we establish a risky driver recognition model as a case study, using video-extracted vehicle trajectory data of 2427 private cars on a German highway. Based on rear-end collision risk evaluation, only 4.29% of all drivers are labeled as risky drivers. The inputs of the recognition model are the discrete Fourier transform coefficients of target vehicle's longitudinal speed, lateral speed, and the gap between the target vehicle and its preceding vehicle. Among 12 sampling methods, 2 cost-sensitive loss functions, and 2 probability calibration methods, the result of automated machine learning is consistent with manual searching but much more computation-efficient. We find that the combination of Support Vector Machine-based Synthetic Minority Oversampling TEchnique (SVMSMOTE) sampling, cost-sensitive cross-entropy loss function, and isotonic regression can significantly improve the recognition ability and reduce the error of predicted probability.

Safety evaluation model of urban cross-river tunnel based on driving simulation.

Currently, Shanghai urban cross-river tunnels have three principal characteristics: increased traffic, a high accident rate and rapidly developing construction. Because of their complex geographic and hydrological characteristics, the alignment conditions in urban cross-river tunnels are more complicated than in highway tunnels, so a safety evaluation of urban cross-river tunnels is necessary to suggest follow-up construction and changes in operational management. A driving risk index (DRI) for urban cross-river tunnels was proposed in this study. An index system was also constructed, combining eight factors derived from the output of a driving simulator regarding three aspects of risk due to following, lateral accidents and driver workload. Analytic hierarchy process methods and expert marking and normalization processing were applied to construct a mathematical model for the DRI. The driving simulator was used to simulate 12 Shanghai urban cross-river tunnels and a relationship was obtained between the DRI for the tunnels and the corresponding accident rate (AR) via a regression analysis. The regression analysis results showed that the relationship between the DRI and the AR mapped to an exponential function with a high degree of fit. In the absence of detailed accident data, a safety evaluation model based on factors derived from a driving simulation can effectively assess the driving risk in urban cross-river tunnels constructed or in design.

Hit-and-run crashes in urban river-crossing road tunnels.

Hit-and-run crashes are a relatively infrequent but severe offense worldwide because the identification and emergency rescue of victims is delayed, which increases the injury severities and the mortality rate. However, no studies have been conducted on hit-and-run crashes in urban river-crossing road tunnels (URCRTs), which can greatly threaten the safety of motorists driving in the tunnels. This study, which employs a dataset of vehicle crashes that happened in thirteen urban road tunnels traversing the Huangpu River, established a binary logistic regression model to identify thirteen factors that contribute to escaping after crashes in Shanghai related to the offending drivers, the vehicular and environmental conditions, the tunnel characteristics and crash information. Among the thirty-five variables considered, this study found that a perpetrator's tendency to leave the crash scene without reporting an accident was higher at night, in the tunnel exit, near to or in short tunnels, when a two-wheeled vehicle or heavy goods vehicle (HGV) was involved and when alcohol was involved. While a perpetrator was more likely to remain on the scene in the tunnel entrance, on a rainy day, in a rear end collision, when a bus was involved, in a single vehicle or a multi-vehicle accident. Based on these findings, several countermeasures for better supervision and hit-and-run prevention are proposed.

Risk factors affecting the severity of traffic accidents at Shanghai river-crossing tunnel.

OBJECTIVE: With increasing traffic volume and urban development, increasing numbers of underground tunnels have been constructed to relieve conflict between strained land and heavy traffic. However, as more long tunnels are constructed, tunnel traffic safety is becoming increasingly serious. Thus, it is necessary to acquire their implications and impacts. This study examined 4,539 traffic accidents that have occurred in 14 Shanghai river-crossing tunnels for the period 2011-2012 and analyze the correlation between potential factors and accident injury severity. METHODS: An ordered logit model was developed to examine the correlation between potential factors and accident injury severity. RESULTS: Results show that increased injury severity is associated with male drivers, drivers aged 65 years or older, accident time from midnight to dawn, weekends, wet road surface, goods vehicles, 3 or more vehicles, 4 or more lanes, middle speed limits (50-79 km/h), zone 3, extra-long tunnels (over 3,000 m), and maximum longitudinal gradient. CONCLUSIONS: This article aims to provide useful information for engineers to develop interventions and countermeasures to improve tunnel safety in China.

A new access density definition and its correlation with crash rates by microscopic traffic simulation method.

Better access management can improve highway safety by reducing potential crashes and conflicts. To make adequate access management decisions, it is essential to understand the impact of different access types on roadway safety, usually represented by the crash rate of a roadway segment. The objective of this paper is to propose a new access density definition reflecting the impact of traffic speed variation of different access types. The traffic speed variation was obtained from a microscopic traffic simulation software package TSIS-CORSIM. A sample roadway Temple Terrace Highway was selected to perform traffic simulation. Access Weight was obtained from traffic speed variation, and access density was obtained from access weight. The proposed access density was then compared with the existing definition by analyzing their correlations with crash rates on one suburban street in Temple Terrace, Florida. The comparison demonstrates that crash rates are more highly correlated with the proposed access density than that in the previous study, which is helpful for Federal Highway Administration (FHWA), United States Department of Transportation (USDOT), and transportation consulting companies to regulate the construction, management and design of roadway segments.

Evaluating the safety impacts of the number and arrangement of lanes on freeway exit ramps.

The primary objective of this study is to evaluate the impacts of the number and arrangement of lanes on freeway exit ramps on the safety performance of freeway diverge areas. The research team collected crash data at 343 freeway segments in the state of Florida. Four different types of exit ramps were considered in this study. They were defined as type 1, type 2, type 3, and type 4 exit ramps respectively. Cross-sectional comparison was conducted for comparing crash frequency, crash rate and crash severity between different types of freeway exit ramps. Crash prediction models were developed to identify the factors that contribute to the crashes reported at selected freeway segments and to provide quantified information regarding the safety impacts of different freeway exit ramps. It was found that the ramp and freeway AADT, posted speed limit on freeway, deceleration lane length, right shoulder width, and the type of exit ramp significantly affected the safety performance of freeway diverge areas. The study demonstrated the safety benefits of using lane-balanced exit ramps. Based on the crash prediction models, replacing a type 1 exit ramp (lane-balanced) with a type 2 exit ramp (not lane-balanced) will increase crash counts at freeway diverge areas by 68.33%. Replacing a type 3 ramp (lane-balanced) with a type 4 ramp (not lane-balanced) will increase crash counts at freeway diverge areas by 32.20%.

Safety effects of the separation distances between driveway exits and downstream U-turn locations.

Using U-turns as alternatives to direct left-turns is an important access management treatment which has been widely implemented in the United States to improve safety on multilane highways. The primary objective of this study is to evaluate the safety effects of the separation distances between driveway exits and downstream U-turn locations. To achieve the research objective, crash data reported at 140 street segments in the state of Florida were investigated. The selected sites were divided into three groups based on the separation distances. t-Tests and proportionality tests were performed for comparing crash frequency, crash type, and crash severity between different separation distance groups. Negative-binomial models were developed for examining the factors that contribute to the crashes reported at selected sites. The data analysis results show that the separation distances significantly impact the safety of the street segments between driveways and downstream U-turn locations. A 10% increase in separation distance will result in a 3.3% decrease in total crashes and a 4.5% decrease in the crashes which is related with right-turns followed by U-turns. The models also show that providing U-turns at a signalized intersection will result in more crashes at weaving sections. Thus, if U-turns are to be provided at a signalized intersection, a longer separation distance shall be provided.

Factors influential in making an injury severity difference to older drivers involved in fixed object-passenger car crashes.

To identify factors influencing severity of injury to older drivers in fixed object-passenger car crashes, two sets of sequential binary logistic regression models were developed. The dependent variable in one set of models was driver injury severity, whereas for the other it was the crash severity (most severe injury in the crash). For each set of models, crash or injury severity was varied from the least severity level (no injury) to the highest severity level (fatality) and vice versa. The source of data was police crash reports from the state of Florida. The model with the best fitting and highest predictive capability was used to identify the influence of roadway, environmental, vehicle, and driver related factors on severity. Travel speed, restraint device usage, point of impact, use of alcohol and drugs, personal condition, gender, whether the driver is at fault, urban/rural nature and grade/curve existence at the crash location were identified as the important factors for making an injury severity difference to older drivers involved in single vehicle crashes.