Understand Driver Awareness through Brake Behavior Analysis: Reactive versus Intended Hard Brake

Understand Driver Awareness through Brake Behavior Analysis: Reactive versus Intended Hard Brake



​​​Driving is a highly dynamic activity where drivers’ awareness to the traffic environment plays the essential role for successful performance. Easy, smooth driving depends on drivers’ awareness to develop situation-specific expectations, where infrequent or unexpected situations are not taken into account. Understanding these unexpected situations provides important insight on driver situation awareness and accident prevention. This study takes advantage of recent development of wearable devices and uses driver physiological signals to identify such ​unexpected situations during driver hard brake.  Based on a naturalistic driving dataset, we define two types of hard brake behavior: reactive and intended hard brake. The reactive hard brake relates to drivers reacting to unexpected situations that usually leads to deviated physiological signals due to stress. The intended hard brake relates to planned maneuver implementation that consists of stable physiological signals. By using the human evaluation, we identified the different situations in which these two types of hard brake occurs. Clear difference is observed between these two groups of situations. Our goal is to identify features that are representative of these two type of road environment, especially the situations where unexpected reactive hard brake happens. Following this direction, we extracted features from Lidar depth scanner to represent the road scene, and applied both lasso regression and logistic regression classifier for feature analysis. The regression model achieves high correlation of 0.77 between the prediction and the ground truth while the classification achieves Fscore of 0.76. The selected Lidar features can serve as high level road scene representation that facilitate next generation advanced driver assistance systems (ADAS) to prevent accident in unexpected traffic scenarios.


Proceedings of Intelligent Transportation Systems (ITSC)
11 Jun 2017
N. Li, T. Misu, F. Tao