Data-Driven Applications for Connected Vehicle Based Traffic Signal Systems

Abstract

Massive deployment of connected vehicles (CVs) is now on the horizon, and will undoubtedly introduce paradigm shifts to the transportation system. At signalized intersections, CV can receive real-time traffic information from roadside equipment (RSE) so that driver can be advised for safer driving, while signal controllers can receive vehicle position and speed information for more effective operation. Considering signalized intersections are often hot-spots of traffic congestion and driver frustration, tremendous opportunities exist to improve the effectiveness and efficiency of traffic signal operation with CV data. However, due to the lack of real-world CVs, the benefit of CV data for signal operation has yet been explored. This limitation has now been partially overcome with the safety pilot model deployment (SPMD) project, the world's first large-scale CV deployment project with around 2,800 CVs.

Through leveraging the SPMD project, this dissertation is the first-ever effort of analyzing large amount real-world CV data to improve traffic signal system operation. Three innovative traffic signal applications are developed to explore the benefit of CV data with low penetration rates. Firstly, to facilitate the deployment of Vehicle-to-Infrastructure (V2I) systems at intersections, a procedure is developed for automatic generation of intersection map, a critical element of many CV applications. Using data from RSE, the proposed approach can automatically estimate intersection geometry and lane-phase mapping, and serve as a cost-effective alternative to prepare input for RSE deployment. Secondly, to pave the way for detector-free signal operation, an algorithm is developed for estimating traffic volumes with CV data. This application could help reduce the dependency of traffic signals on vehicle detectors, and would be particularly beneficial for signal operation. Lastly, to explore the benefit of V2I communication for driving assistance, a speed advisory system is proposed to help drivers reduce fuel consumption when driving through intersections, using information from RSEs. An efficient algorithm is proposed based on Pontryagin's maximum principle for real time implementation.

With the three applications to improve traffic signal system in three different perspectives, the ultimate objective of this dissertation is to facilitate development and deployment of CV-based traffic signal system in the near future.