01661270

Component

This study introduces a crash/near-crash prediction model developed from vehicle kinematics data. The study hypothesis is that vehicles experience significant turbulence in their kinematics before involvement in crashes/near-crashes. To test this hypothesis, the SHRP2 NDS vehicle kinematics data (speed, longitudinal acceleration, lateral acceleration, yaw rate, and pedal position) are utilized. The data are first prepared based on two approaches: Euclidean point and similarity matrix. In the first approach, several algorithms are trained and comparatively analyzed including K Nearest Neighbor (KNN), Random Forest, Support Vector Machine (SVM), Decision Trees, Gaussian Neighborhood, Adaptive Boost (AdaBoost), Multilayer Perceptron (MLP), and Quadratic Discrimination Analysis (QDA), whereas the kernel SVM algorithm is tested in the second approach. Initial testing indicates that AdaBoost outperforms all other methods in the Euclidean point approach. Sensitivity analysis is accordingly performed using AdaBoost and the kernel SVM models to determine the optimal prediction horizon length (the time period before which a crash/near-crash can be predicted) and turbulence horizon length (the time period over which crash/near-crash related changes in vehicle kinematics take place). The results reveal that both models have considerably reliable prediction accuracy around 90% at one-second prediction horizon and four-second turbulence horizon. It is consequently believed that such time window allows for capturing the crash/near-crash related variations in vehicle kinematics. The achieved high prediction accuracy is promising for crash avoidance systems in the emerging autonomous vehicle technology.

This paper was sponsored by TRB committee ANB20 Standing Committee on Safety Data, Analysis and Evaluation.

18-03927

English

7p

2018

Transportation Research Board 97th Annual Meeting

Digital/other

Figures; References (14)

Automatic data collection systems; Crash avoidance systems; Crash data; Mathematical models; Mathematical prediction; Vehicle dynamics

Data and Information Technology; Highways; Safety and Human Factors

Transportation Research Board Annual Meeting 2018 Paper #18-03927

TRIS, TRB, ATRI

Jan 8 2018 10:58AM