Developing Agent-Based Distributed Cooperative Vehicle-Infrastructure Systems in the Connected and Automated Vehicle Environment

The rapid development of our transportation systems has brought much convenience to our daily lives, while also introducing various issues related to safety, mobility, and environmental sustainability. To address these transportation system issues, connected and automated vehicle (CAV) technology has had significant development during the last decade. With CAV technology, the capabilities of vehicles is greatly improved, allowing for equipped vehicles to not only drive partially or fully automatically using information from on board sensors, but also behaving cooperatively through vehicle-to-everything (V2X) communications.

In this dissertation research, agent-based distributed cooperative vehicle-infrastructure systems have been developed to evaluate CAV applications from the perspective of safety, mobility and environmental sustainability, both qualitatively and quantitatively. Specifically, the proposed CAV systems described in this dissertation are divided into two major categories: The first category is based only on vehicle-to-vehicle (V2V) communication among CAVs, while the second category is based on both V2V and infrastructure-to-vehicle (I2V) communication.

In this dissertation, three different cooperative automation applications are addressed, including cooperative adaptive cruise control (CACC), cooperative eco-driving at signalized intersections, and cooperative merging at highway on-ramps. Different agent-based motion control algorithms of CAVs are proposed and evaluated, including a distributed consensus control algorithm, an online feedforward/feedback control algorithm, and an optimal control algorithm. All proposed applications are qualitatively and quantitatively evaluated using numerical simulations and/or microscopic traffic simulations, showing their benefits of avoiding collisions, increasing traffic flow, decreasing travel time, and/or decreasing energy consumption and pollutant emissions. Field implementations with real vehicles traveling in the real-world traffic environment have also been conducted to evaluate the effectiveness of the proposed algorithms.

Additionally, studies of the aforementioned cooperative automation applications were also carried out using a game engine, which provided a simulation environment with more realistic vehicle models and road networks. Human-in-the-loop simulations were conducted on the driving simulator platform, and a learning-based approach was developed to model the human factors.