Prioritize Winter Crash Severity Influencing Factors in US Midwestern for Autonomous Vehicle
Abstract
Adverse weather conditions in winter have significant impacts on crash occurrences and risks. Human drivers can adjust driving styles based on the context information of the surrounding traffic and environments. Similar schemes should be considered and designed into autonomous vehicle (AV) control systems. However, most of the existing autonomous vehicle control systems do not have effective mechanisms to deal with extreme weather conditions. There are very limited numbers of research works have focused on the risk factors influencing crash severity affected by winter precipitation. In this study, we aim to find out how different weather conditions relate to crash severities and what are the most influencing risk factors for autonomous vehicles. We utilized three-year crash data of the state from Wisconsin for this study. We evaluated the performance of three statistical prediction models and compared the importance of relevant factors with all crashes and crashes affected by winter precipitation. Evaluation results showed that different weather conditions have a significant influence on crash risk factors. Finally, we prepared a prioritized list of variables that has potential significant impacts on autonomous vehicles safety. Our findings might be useful for designing the control system to improve AV safety under adverse weather conditions.
Subject
crash severity
winter precipitation
machine learning
transportation safety
autonomous vehicles