Data-Driven Inclusive Framework for Safety Assessment of Connected and Autonomous Vehicular Traffic

Abstract

Intersections are planned areas within the roadway infrastructure where conflicts between various roadway users may occur. Various strategies, including traffic signals, markings, and roadway signs, control traffic movements along the potential conflict areas. Despite these control strategies, crashes between various road users within intersections have historically represented a large portion of overall roadway crashes. Crashes between road users in other safety-critical areas within the roadway infrastructure such as high-speed freeways also significantly contribute to overall crashes. Therefore, the assessment of road traffic safety for its improvement has been a widely pursued research topic.A critical challenge that current traffic safety assessment methods face is the lack of high-fidelity data explaining the pre-crash sequence of interactions between various road users. Recent advances in vehicle-to-vehicle and vehicle-to-infrastructure communication technology have enabled the availability of new sources of high-fidelity and information-rich data. Connected vehicles broadcast their dynamic state information and connected intersections broadcast static information about the geometry of their critical physical features along with their dynamic signal state information. Industrywide efforts are underway to facilitate additional information sharing between infrastructure and vehicles, such as object locations, weather conditions, road conditions, and dynamic situational information. Technology-driven advances in infrastructure and modern vehicles provide an ability to capture the information-rich data that can be used to assess road traffic safety. When combined, the information from these new data sources brings an opportunity to rethink current methods of traffic safety assessment, develop new methods by combining classical approaches with modern approaches, and improve the overall situational awareness of both vehicles and infrastructure. This dissertation presents a framework of a scalable system developed to capture information from new data sources and modern intersection traffic safety assessment methods that are built upon traditional conflicts-based surrogate safety measures. The presented field data capture system is deployed on multiple live intersections in different geographical locations within the United States. One of the challenges in the framework development is the currently limited availability of on-road connected vehicles. To overcome this challenge, this dissertation presents a simulation-based data capture framework as a source of the vehicle and infrastructure data structurally identical to the data collected from the field. Safety assessment methods presented in this dissertation utilize the information from newly available data sources to reveal the probabilistic nature of safety-critical zones within intersections and other roadway areas in both space and time. The impacts of measurement uncertainties on safety assessment methods are analyzed. The end goals of this research are to (1) enable transportation agencies to make informed decisions towards improving the intersection traffic safety, and to (2) improve the situational awareness of drivers of communication enabled vehicles to potentially reduce safety risks by providing information about safety critical areas of the intersection as they navigate through the intersection and other roadway areas.