Statistical Model of Nationwide Crash Data Assessed Safety Benefits for Level 1 and 2 Automated Vehicles.
Nationwide, United States
Influence of Level 1 and Level 2 Automated Vehicles on Fatal Crashes and Fatal Crash Occurrence
Summary Information
Connected and Automated Vehicles (CAV) are expected to improve safety by gradually reducing human errors while driving through the integration of smart features or by fully eliminating human driving roles at higher levels of vehicle automation. This study aimed to identify risk factors influencing fatal crashes involving level 1 (driver assistance) and level 2 (partial driving automation) CAVs in the USA and evaluate the effectiveness of various smart features in improving safety to understand the overall safety effectiveness of CAVs. Fatal crash data from 2016 to 2019 were obtained from the Fatality Analysis Reporting System (FARS) database and smart features in each vehicle involved in a crash were retrieved from the National Highway Traffic Safety Administration (NHTSA) database. Fatal crashes involving level 0 vehicles (non-connected and automated vehicles) were compared to minimize unobserved heterogeneity and randomness associated with the influencing risk factors. In addition, vehicles with smart features (e.g., pedestrian automatic emergency braking [PAEB] and lane keeping assistance [LKA]) and other vehicles were compared in terms of their safety performance.
METHODOLOGY
A partial proportionality odds model was developed to examine the relationship between crash, road, and vehicle characteristics as the independent variables and the occurrence of a fatal crash involving a vehicle with a specific level of automation as the dependent variable. The odds ratio for each parameter represents the odds of a specific outcome compared to the other outcomes. To minimize the spatial variance in the geometric characteristics, for every crash involving a level 1 or level 2 vehicle, three nearest fatal crashes involving level 0 (non-automation) vehicles were identified as corresponding samples for level 0 crashes in modeling. In the analysis, samples involving one or more CAVs with varying levels of automation were also identified and treated as distinct data points. To evaluate the efficacy of various smart features, a comparative analysis was conducted between vehicles equipped with features and other vehicles involved in fatal crashes.
FINDINGS
- Level 1 and level 2 CAVs were found to be less likely to be involved in a fatal crash at four-way intersections, on two-way routes with wide medians, at nighttime (12:00 am – 2:59 am), at railway grade crossings, and due to other vehicle’s loss of control when compared to level 0 vehicles.
- However, level 1 and 2 CAVs were found to be more likely than level 0 vehicles to be involved in a fatal crash with pedestrians and bicyclists attributed to decreased attentiveness of drivers due to reliance on technology. Also, the chances of a fatal crash involving a CAV at a flashing signal or stop sign was also higher than level 0 vehicles. Pedestrians were identified as the most vulnerable road users.
- Furthermore, comparative analysis between vehicles with smart features and other vehicles indicated that PAEB and LKA improve the safety by reducing possible collision with a pedestrian and roadside departure. The proportion of pedestrian crashes for vehicles equipped with a PAEB system (16.35 percent) was less than the corresponding proportion from all fatal crashes (19.32 percent). Similarly, the proportion of roadside departure crashes for vehicles equipped with a LKA system (10.12 percent) was less than the corresponding proportion from all fatal crashes (14.51 percent).
- However, adaptive cruise control (ACC) and forward collision warning system (FCWS) are not efficient in improving safety in case of rear-end collisions. The proportion of rear-end collisions for vehicles equipped with ACC and FCWS (14.69 and 14.71 percent) was higher than the corresponding proportion from all fatal crashes (7.07 percent), respectively.
