Assessing Surrogate Safety Measures Using a Safety Pilot Method Deployment (SPMD) Dataset Zhaoxiang He, Ph.D. Student Xiao Qin, Associate Professor Md Abu Sayed, Ph.D. Student Department of Civil and Environmental Engineering University of Wisconsin-Milwaukee

Introduction Crash history has long been considered by safety professionals as a principal and straightforward performance measure for road safety; however, its limitations suggest more robust surrogate safety measures. Emerging data sources such as Safety Pilot Model Deployment (SPMD) provide a great opportunity to gain a better understanding of collision mechanisms and to develop novel safety metrics. In-vehicle data (e.g., speed, location) collected by the SPMD program can be an important supplement to traditional crash data oriented safety analysis. The goal of this study is to retrieve proper information from vehicle trajectory data in SPMD, construct several surrogate safety measures, and assess their performance.

Data Collection Only rear-end crashes in the mid-block were considered in this study. Road network data and crash data (2011-2015) for Washtenaw County were collected from the Southeast Michigan Council of Governments Open Data Portal. Two months of the SPMD dataset (Oct. 2012 and Apr. 2013) are free to download in Research Data Exchange (RDE). This study used the DataWsu file (GPS-based data such as speed and longitudinal acceleration, and vehicle states such as brake status) and the DataFrontTargets file (front target information such as distance to front target and relative speed for the front target) in the driving dataset DAS1 collected from around 100 instrumented vehicles.

Data Collection

Methodology *VF is the front vehicle speed, VS is the subject vehicle speed, and the relative speed ΔV is (VS - VF). aF is the front vehicle acceleration, aS is the subject vehicle acceleration, and the relative acceleration Δa is (aS - aF). D is the distance between two vehicles. Vehicle kinematics from the SPMD data set were used to calculate vehicle-level safety surrogate measures at each timestamp, and then aggregated into trip-level safety surrogate measures for each link. The link-level indexes aggregate trip-level indexes on the link.

Safety Surrogate Measures

Safety Surrogate Measures

Results The statistical relationships between surrogate safety measures and rear-end crashes were established using negative binomial (NB) models. Additional safety information was included, such as vehicle maneuvering decisions (e.g., average speed, brake duration) and link length. Model results showed that modified time to collision (MTTC) outperforms time to collision (TTC) and deceleration rate to avoid collision (DRAC) in terms of goodness of fit.

Negative Binomial Models for Mid-block Rear-end Crashes

Conclusions In this study, a SPMD dataset was used to develop and evaluate surrogate safety measures for mid-block rear-end crashes. Some of the assumptions or data processing approaches used in this study may not be optimal in all situations; thus, future studies should search for more efficient ways. Future studies can be expanded into the comparison of other safety surrogate measures such as post encroachment time (PET). New and emerging safety surrogate measures can be developed with the rich information provided through connected vehicle safety technologies.

Thank you ! Comments?