An Exploratory Approach to Digitizing the Operational Environments for Connected and Automated Vehicles

The transportation industry is going beyond the traditional areas. The digital innovations operated by the technology companies are reinventing transportation with new business and technologies, such as shared mobility and autonomous driving. Among these, connected and autonomous vehicles (CAV) emerge to be a strike in both the industry and research areas. However, in comparison to the thriving in public discussion, the practice and the market of autonomous driving are not clear yet. To ensure operation safety, the CAVs are equipped with various sensors to perceive the surrounding environments during the operation. The road tests of the CAVs collect comprehensive driving contextual information which is not available in the past. The data conveys the information about the vehicle operation, interaction, and the static road environments. The booming in the CAV data provides extensive sources to investigate the traffic from a microscopic aspect, which brings an evolution to the traditional transportation industry and academics to transform from the planning, design, and operation to the age of data, modeling, and machine learning. Under this trend, this dissertation takes the advantage of the multi-source data to examine the driving environments that the CAVs will encounter. The overarching research goal of this dissertation is to explore a framework or methodology to understand the driving environments from the view of CAVs. This research utilizes the (1) Connected vehicle basic safety message data; (2) Google street image data; (3) Lyft Level 5 perception data and (4) Waymo motion data to explore the driving environments from both the static and dynamic aspects. The methodology of this research incorporates spatiotemporal analysis, statistical modeling, and machine learning. The dissertation research will be unfolded into 4 sections which are targeted at four datasets: (1) Historical driving performance study; (2) Static driving environment classification; (3) Dynamic driving environment characterization; (4) Contextual vehicle lane changing prediction.The research in this dissertation is expected to contribute by providing practical metrics for digitizing the driving environments. The extracted information can be compiled to the High Definition (HD) map for autonomous driving or can be employed as indicators for the Operational Design Domains (ODD) evaluation.