Connected Vehicle Data Platform

Related Publications

Overview

My early PhD work focused on building a connected vehicle (CV) data platform to support advanced CV applications. At the time, the Ann Arbor Connected Vehicle Test Environment (AACVTE) was the world’s largest real-world deployment of connected vehicles, featuring over 2,500 vehicles and 74 infrastructure sites. A significant amount of data from both vehicles and infrastructure was continuously collected in real time. I developed the AACVTE data platform almost single-handedly, managing data reception, preprocessing, and visualization. The platform processed streaming data in real time, integrating vehicle data, infrastructure information (e.g., signal phase and timing), and road networks.

Following the AACVTE project, we collaborated with GM to utilize their connected vehicle data, which provided more complete trip information, from origin to destination. I upgraded the data platform to support larger-scale applications, enabling it to automatically generate network-level performance measurements. This platform, which remains under maintenance, has facilitated numerous CV-based applications. Beyond traffic signal control systems which will be discussed later, it also served as the foundation for various research projects at the University of Michigan’s transportation program, including dilemma zone detection at signalized intersections, inference of signal timing data, driving behavior classification, and large-scale network model calibration, etc.

Our Python library for this data platform: MTLDP. Due to license issue, this library is now private.