Rapid Driving Style Recognition in Car-Following Using Machine Learning and Vehicle Trajectory Data

Abstract

Rear-end collision crash is one of the most common accidents on the road. Accurate driving style recognition considering rear-end collision risk is crucial to design useful driver assistance systems and vehicle control systems. The purpose of this study is to develop a driving style recognition method based on vehicle trajectory data extracted from the surveillance video. First, three rear-end collision surrogates, Inversed Time to Collision (ITTC), Time-Headway (THW), and Modified Margin to Collision (MMTC), are selected to evaluate the collision risk level of vehicle trajectory for each driver. The driving style of each driver in training data is labelled based on their collision risk level using K-mean algorithm. Then, the driving style recognition model’s inputs are extracted from vehicle trajectory features, including acceleration, relative speed, and relative distance, using Discrete Fourier Transform (DFT), Discrete Wavelet Transform (DWT), and statistical method to facilitate the driving style recognition. Finally, Supporting Vector Machine (SVM) is applied to recognize driving style based on the labelled data. The performance of Random Forest (RF), K-Nearest Neighbor (KNN), and Multi-Layer Perceptron (MLP) is also compared with SVM. The results show that SVM overperforms others with 91.7% accuracy with DWT feature extraction method.

A Systematic Methodology to Evaluate Prediction Models for Driving Style Classification

“…This result is consistent with previous studies reported by Xue et al [18] and Wang et al [17]. In Reference [18], authors indicated that the SVM model (accuracy = 0.917) outperformed RF, kNN, and ANN models using acceleration, relative distance, and relative velocity features for classifying normal and aggressive driving styles from 320 driving tests in a naturalistic setting. Similarly, Wang et al [17] demonstrated that the SVM model with RBF kernel (accuracy = 0.772) was also better than a linear kernel (accuracy = 0.86) using speed and throttle opening features to classify normal and aggressive driving styles.…”

“…Dörr et al utilized three simulated virtual scenarios to label their dataset as comfortable, normal, or sporty driving styles [15]. Another set of studies reported the use of unsupervised machine-learning models, such as K-means, to find common group patterns that can be related to driving styles, such as calm or aggressive [13] and normal or aggressive [17,18].…”

“…In addition, in the area of advanced driver-assistance systems (ADAS), researchers are interested in the development of intelligent tools to provide tailored feedback according to the driver's behavior [11].Current studies have reported different methodologies to classify driving styles using in-vehicle data. These methodologies mostly depend on (i) the inputs that can be extracted and derived from the data gathered (e.g., acceleration, deceleration, brake) [12,13]; (ii) the computational models to classify driving styles [14,15]; (iii) driving-styles outputs (e.g., calm, normal, aggressive) [16,17]; and (iv) the performance metrics that evaluate these models [18,19]. Although these studies show a growing interest in classifying driving styles empirically, few studies have systematically evaluated which inputs and models are better predictors of a particular driving style's output.The existing body of research has attempted to evaluate several computational models to identify driving styles using well-established metrics, for example, accuracy, precision, and recall.…”

“…Although these studies show a growing interest in classifying driving styles empirically, few studies have systematically evaluated which inputs and models are better predictors of a particular driving style's output.The existing body of research has attempted to evaluate several computational models to identify driving styles using well-established metrics, for example, accuracy, precision, and recall. These studies have reported that Random Forests (RF), Artificial Neural Networks (ANN), k-Nearest Neighbors (kNN), and Support Vector Machine (SVM) are the best classifiers for estimating driver events [14], driving-style maneuvers [19], and driver's aggressiveness [18]. However, few empirical works have attempted to select the best model across different driving-styles classifiers based on reviewed literature and current research in the field.A review conducted by Meiring and Myburgh [20] suggested that fuzzy logic and SVM can be considered the most promising simple classifiers (c.f.…”

“…The existing body of research has attempted to evaluate several computational models to identify driving styles using well-established metrics, for example, accuracy, precision, and recall. These studies have reported that Random Forests (RF), Artificial Neural Networks (ANN), k-Nearest Neighbors (kNN), and Support Vector Machine (SVM) are the best classifiers for estimating driver events [14], driving-style maneuvers [19], and driver's aggressiveness [18]. However, few empirical works have attempted to select the best model across different driving-styles classifiers based on reviewed literature and current research in the field.…”