UC Berkeley

Connected Vehicle data is an emerging technology that offers to broadly measure traffic at ascope not accessible with traditional traffic measurement systems. Using GPS and vehicles onboard data, we can gain a broader perspective of traffic. The current adoption rate of connected vehicles is near 2-3 percent, but offers wide-area coverage. The resulting sparsity of data offers new challenges in extracting useful measures of the traffic state. This work focuses on an empirical study of available connected vehicle data collected from the San Francisco and Los Angeles areas. We explore how the penetration rate of connected vehi- cles in these areas influence the fidelity of extracted traffic flow measurements. Leveraging the accuracy of the GPS samples, we explore separating data into lane-level distinctions. We developed penetration rate agnostic estimates of queue length distributions at intersec- tions. Utilizing the repeatability of timed intersections, we derive spatiotemporal diagrams of a major throughway and connect them to a macroscopic fundamental diagram. Finally, we explore new accident risk calculations based on extracted maneuver level information. Connected vehicle data offers new and exciting insights that will continue to improve with increased penetration rate.