Research Papers

A Perceptual Forward Collision Warning Model using Naturalistic Driving Data

Version 1
Date added July 10, 2018
Downloaded 0 times/fois
Category 2018 CARSP XXVIII Victoria
Tags Research and Evaluation, Session 5B
Author/Auteur H. Tawfeek, El-Basyouny
Stream/Volet Research and Evaluation

Slidedeck Presentation Only (no paper submitted)

5B - Tawfeek - A Perceptual...

Abstract

Rear-end collisions represent a quarter to one-third of the total number of collisions occurring on North American roads. While there are several methods to mitigate rear-end collisions effects, one way is to warn drivers about impending events using Forward Collision Warning (FCW) systems. At the core of any FCW algorithm is a trigger distance at which a message is relayed to the driver to avoid rear-end collisions. The main goal of this paper is to propose a warning distance model based on naturalistic driver following behavior. The goal of this paper was achieved by investigating car-following events within a critical time-to-collision (TTC) range. A total of 5785 candidate car-following events were identified in Naturalistic Driving Study data for the model development. Using regression analysis, the minimum warning distance was linked to several performance measures.

To assess the performance of the developed algorithm, it was compared to six of the existing FCW algorithms in terms of warning distances. These six algorithms are grouped in two main classes; namely, perceptual and kinematic algorithms. The former algorithms consider the perceptual ability as a priority to address rear-end collisions. On the other hand, the kinematic algorithms group were derived based on vehicle kinematics with various assumptions on accelerations. The comparison between the algorithms was conducted in terms of comparing the warning distance and the relative speed at various following vehicle speeds (i.e., 60, 80, 100, and 120 km/h). For the development of the warning distance model, it was found that the relative speed, the host vehicle speed, and the host vehicle acceleration can significantly affect the minimum warning distance. The model had a R-Squared value of 0.916 indicating that the model provided a good fit to the data.

For the model assessment, the comparison revealed that the warning distances varied widely between algorithms with perceptual algorithms having shorter distances compared to kinematic algorithms. The results of the developed algorithm were consistent with the other perceptual algorithms. However, the proposed algorithm had longer warning distance when compared to Algorithm 1 for relative speeds higher than 10 km/h at medium host vehicle speeds (i.e. 60 and 80 km/h) and at high host vehicle speeds (i.e. 100 and 120 km/h). Otherwise, the developed algorithm, in most of the cases, had shorter warning distance than the other five algorithms. In general, the distance produced by the proposed algorithm is less than the distance produced by the perceptual algorithms which were developed based on driving simulator data. This means that the FCW developed using driving simulators will unnecessarily warn drivers earlier. Such early warnings reduce the FCW systems acceptance and decreases the warning impact on driver behavior.

Therefore, this finding highlights the importance of developing warnings based on actual behavior. The warning distance produced by the developed algorithm was less than the other algorithms which were developed using driving simulator or test tracks. This paper integrated drivers' actual behavior in the development of FCW algorithm, and hence, it is expected to be more acceptable by drivers.