Abstracts Transportation

Improving Road Safety Through Detecting Invisible Area and Collaborative Perception

by Yongxin Wang

Safety assurance is one of the basic issues in modern transportation. According to the World Health Organization(WHO), there were 1.35 million deaths attributed to road traffic crashes worldwide each year. A recent vehicle crash causation study by NHTSA found that human errors are the critical reasons in 94\% of the crashes. The emergence of autonomous driving systems and Vehicle-to-Everything (V2X) communication could drastically reduce the number of crashes and fatalities that occur on the roads today. This dissertation focuses on improving the safety in transportation systems by doing so with connected and autonomous vehicles. The first part of this dissertation examines how to predict the failure cases of the perception systems in autonomous vehicles. Knowing when the perception system will fail is a necessity to find alternate control methods so that crashes can be mitigated for both self-driving and human driven vehicles. The second part of this work addresses the viability of using collaborative perception to mitigate potential road risks by enabling Non-Line-Of-Sight (NLOS) obstacle detection. While the majority of previous research is based on simulation, I have conducted experiments using on-board devices in real road going vehicles in real traffic scenarios (not on a test track with controlled traffic) that have communication latency and uncertainty. The third part of the dissertation proposes a method to analyze the safety and security of cyber security systems and studied in detail a critical use case of the movable railroad bridges. In addition, the work reported in this dissertation has created a large corpus of multi-modal multi-weather driving data to facilitate research in perception studies of autonomous and connected vehicles that are used by vehicle controllers.