Road safety evaluations mainly rely on the analysis of crash data that are challenged by well-recognized availability and quality issues. The statistical models used to predict the safety level of road sites—that is, safety performance functions—have recently been successfully developed with the use of traffic conflict observations instead of crashes. As such, it is possible to adopt and transfer the statistical techniques used in crash-based road safety analysis to conflict-based analysis. The use of statistically rigorous techniques in crash-based before-and-after (BA) studies is essential for evaluation of the effectiveness of road safety countermeasures. In particular, the use of Bayesian methods, such as the empirical Bayes (EB) technique, is vital to control for confounding factors that can operate simultaneously with the countermeasure and may affect road safety performance. The main objective of this paper was to estimate the treatment effectiveness of two traffic signal (visibility) improvement projects in the city of Edmonton, Alberta, Canada, with a conflict-based BA study using the comparison group and the EB methods. More than 300 h of video data with traffic conflict observations was automatically collected and analyzed by computer vision techniques for two treatment intersections and two control (untreated) intersections before and after the signal improvement projects. The results of the comparison group method showed a statistically significant 24% reduction in the average number of rear-end conflicts per hour, whereas the EB method showed a statistically significant 24.5% reduction in the average number of total conflicts per hour.

Automated computer vision techniques were used to analyze 2 h of video data collected at a major signalized intersection in New York City. The main objectives of this study were to diagnose pedestrian safety issues and identify contributing factors at the intersection and to demonstrate the feasibility of the automatic extraction of pedestrian data required for pedestrian behavior analysis—mainly pedestrian speed and gait parameters. The safety study was conducted with traffic conflict techniques. The main factor that contributed to the high number of pedestrian and vehicle conflicts was found to be pedestrian violations, mainly temporal violations in which pedestrians crossed the street during the “Don't Walk” or flashing “Don't Walk” phase. During the 2 h analyzed, about one-third of pedestrians were noncompliant with the signal timing or crosswalk boundary (17.9% spatial violations and 15.3% temporal violations). Pedestrian speed, step frequency, and step length were automatically extracted for 333 pedestrians and were found to follow the normal distribution with 95% confidence (mean and standard deviation of 1.47 ± 0.27 m/s, 1.96 ± 0.17 Hz, and 0.75 ± 0.14 m, respectively). Gait analysis showed that the walking speed for single pedestrians was 9% higher than for those who walked in groups. Males tended to be slightly faster than females, with higher step length but lower step frequency. Violators tended to have higher walking speeds compared with non-violators, and the difference in speed was dependent on step length but not on step frequency.

No abstract

No abstract

Over the past few decades, numerous adaptive traffic signal control (ATSC) algorithms have been proposed to alleviate traffic congestion and optimize traffic mobility using real-time traffic data, such as data from connected vehicles (CVs). However, most of the existing ATSC algorithms do not consider optimizing traffic safety, likely because of the lack of tools to evaluate safety in real time. In this paper, we propose a novel ATSC algorithm for real-time safety optimization. The algorithm utilizes a traditional Reinforcement Learning approach (i.e., Q-learning) as well as recently developed extreme value theory (EVT) real-time crash prediction models. The algorithm was validated using real-world traffic video data collected from two signalized intersections in British Columbia. The results indicated that, compared with an existing fully actuated signal controller, the developed algorithm can significantly reduce the real-time crash risk by 43% to 45% at the intersection’s approaches even at low CVs market penetration rates.