Traffic Parameter Methods for Surrogate Safety: A Comparative Study of Three Mobile Sensor Technologies

Author(s): Joshua Stipancic, Luis Miranda-Moreno, Nicolas Saunier

Slidedeck Presentation:

3A - Stipancic


Although maintaining adequate levels of safety is a universal requirement for modern road networks, the preferred techniques for defining and quantifying safety remain debated. Though traditionally popular, crash-based methods are reactive, requiring crashes to occur before causes can be identified. In response, surrogate safety measures have emerged. Existing work in the use of traffic parameters as surrogate measures has predominantly focused on traffic parameter data collected by loop detectors, without comparison of surrogate measures reported by different sensor technologies. The purpose of this paper is to evaluate how three non-intrusive traffic sensors, microwave radar, plate magnetometer, and video-based devices, report safety surrogate measures. The surrogates considered included severe conflicts (measured by time-to-collision, TTC, less than 3 and 5 s), temporal speed variation (measured by the coefficient of variation of speed, CVS), and lateral speed variation (measured by the average difference in speed between adjacent lanes, ΔS). For rear-end TTC, the video-based sensor reported relatively more conflicts than the radar and magnetometer, which performed similarly. CVS calculated from radar data was consistently higher than for the video. These measures are largely influenced by the speed overestimation bias present in video-based data. Utilizing the average difference in speed across lanes to quantify lateral speed variation is independent of mean speed, the overestimation bias is inconsequential for the indicator, and the empirical results from the radar and video detectors are similar as expected. Compared to ground truth data, variability in sensor data was observed to create uncertainty in the surrogate measures for all sensors.