Doctoral Dissertations

Orcid ID

https://orcid.org/0000-0002-9579-9403

Date of Award

5-2023

Degree Type

Dissertation

Degree Name

Doctor of Philosophy

Major

Mechanical Engineering

Major Professor

Zhenbo Wang

Committee Members

Subhadeep Chakraborty, Jindong Tan, Yunli Shao

Abstract

The recent advancements in communication technology, transportation infrastructure, computational techniques, and artificial intelligence are driving a revolution in future transportation systems. Connected and Automated Vehicles (CAVs) are attracting a lot of attention due to their potential to reduce traffic accidents, ease congestion, and improve traffic efficiency. This study focuses on addressing the challenges in controlling future CAV-enabled transportation systems. The aim is to develop a framework for the control of CAV-based traffic systems to improve roadway safety, travel efficiency, and energy efficiency. The study proposes new methods for vehicle speed control and traffic signal control at signalized intersections and corridors as well as merging roadways, to increase the understanding of how traffic elements interact and are impacted by individual actors. The vehicle speed control method is based on sequential convex programming (SCP) algorithms, combining the pseudospectral collocation method with line-search and trust-region techniques for optimal solutions with real-time performance and efficient handling of multiple constraints. In terms of on-ramp merging control, the study develops a new merging control approach that balances computational efficiency, solution optimality, and real-time performance for safe merging operations. The traffic signal control framework uses deep reinforcement learning (DRL) with a novel convolutional autoencoder network for a concise representation of traffic information to improve the learning efficiency of the DRL algorithm. The proposed method extends the action space by including both phase duration and cycle length, allowing for more adaptability to dynamic traffic flow.

This study presents a comprehensive framework for the control of CAV-based traffic system that enhances the positive attributes of CAV technology while minimizing negative effects. The framework will contribute to improving road safety, travel, and energy efficiency while synchronizing CAV motion planning with traffic signal optimization to reduce traffic congestion and idling as well as fuel consumption with guaranteed collision avoidance. This study explores the interface of multiple disciplines including control theory, optimization, machine learning, data analytics, and real-time computation. The results of this study will inform future research in the area of intelligent control of data-rich, interactive systems and will benefit the development of intelligent transportation systems with CAV technologies.

Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS