- 68 views
- 46 downloads
Traffic Management in Mixed Autonomy with CAVs: Sensing, Signal Optimization, and Trajectory Control
-
- Author / Creator
- Wu, Fan
-
Global vehicle numbers continue to climb alongside population growth and economic expansion, resulting in increased traffic congestion, air pollution, and accidents. Major cities worldwide estimate that drivers lose considerable working hours annually to congestion, leading to wasted time, fuel, and increased air pollution levels which impact road users' comfort. Consequently, addressing traffic congestion, reducing vehicle emissions, and ensuring road safety are imperative for sustainable urban development. Intelligent transportation systems (ITS) are vital for sustainable urban development, leveraging emerging technologies such as connected and automated vehicle (CAV) technology, vehicle-to-everything (V2X) communication, and mobile edge computing (MEC). These innovations integrate vehicle automation, real-time communication, and computing resources, offering opportunities for improved traffic management.
Conventional human-driven vehicles (HDVs) and pedestrians are still key players in shaping traffic dynamics today. As such, we must acknowledge the reality of coexistence among various road users, including conventional HDVs, connected vehicles (CVs), CAVs, pedestrians, and cyclists, creating what is termed a 'mixed-autonomy' or 'mixed-traffic' system. However, there is a tendency to overlook the benefits of pedestrians or conventional vehicles when adopting new CAV technology. Therefore, while emerging technologies hold promise, integrating them into real-world transportation scenarios is essential for effectively managing this dynamic traffic system.
Given the complexities of mixed-autonomy traffic and the advancements of CAV technology, this dissertation aims to achieve this through the development of methods for predicting and sensing traffic states, optimizing traffic signal control, and implementing trajectory control approaches for CAVs. The overall goal is to ensure road users’ safety, improve traffic efficiency, reduce vehicular emissions within mixed-autonomy traffic scenarios, and consider the benefits of emerging technologies for all types of road users in mixed traffic. To achieve this goal, the research is structured into three main parts:
The first part develops a precise and efficient method for estimating traffic states using sparse trajectory data from CVs on mixed traffic scenarios involving conventional HDVs and CVs or CAVs, suitable for various road segments such as freeways, highways, or urban arterials. The goal is to predict comprehensive traffic states for both HDVs and CVs, supporting subsequent signal optimization and trajectory control. The study introduces a novel model that utilizes Gaussian processes (GP). By employing a kernel rotation re-parametrization scheme, a standard isotropic GP kernel is transformed into an anisotropic one, allowing for better modeling of traffic wave propagation in flow data. This method effectively estimates traffic states from sparse sensing data collected from fixed sensors, probe vehicles, and CVs. The results demonstrate superior performance of the estimating model in terms of accuracy, efficiency, and robustness.
The second part introduces adaptive signal optimization methods for midblock crossings, tailored for scenarios involving conventional HDVs and pedestrians, as well as scenarios with a mix of HDVs, CVs, and pedestrians. This aims to enhance traffic efficiency along an arterial road, particularly focusing on the urban arterial level with a midblock crossing. It ensures pedestrians' safety while minimizing the impact of frequent crossing requests on the arterial traffic flow. The proposed models utilize the signal control status of adjacent intersections and leverage real-time vehicle location information obtained from CVs to optimize pedestrian waiting time. The optimization model ensures pedestrian safety while enhancing signal coordination between the midblock crossing and downstream intersections. The approach effectively reduces both vehicle and pedestrian delays.
The final section provides an innovative approach to trajectory control specifically tailored for CAVs, with a focus on improving fuel efficiency, safety, and overall traffic performance. This method is developed to handle diverse traffic scenarios involving a combination of HDVs, CAVs, and pedestrians at intersections. The trajectory planning framework integrates a deep reinforcement learning (DRL) algorithm with a multi-agent control strategy. Using the deep deterministic policy gradient (DDPG) algorithm, the proposed method enables CAVs to learn optimal control policies within the complexities of mixed traffic environments. Through thorough evaluations, the proposed framework demonstrates significant enhancements in traffic efficiency, a reduction in vehicle emissions, and an improvement in traffic safety.
The dissertation presents comprehensive research on traffic management within mixed-autonomy traffic environments. The models and algorithms introduced offer effective means to enhance traffic efficiency, minimize vehicle emissions, and prioritize safety. The findings contribute to methodological and practical insights into sustainable and efficient mixed traffic management.
-
- Graduation date
- Fall 2024
-
- Type of Item
- Thesis
-
- Degree
- Doctor of Philosophy
-
- License
- This thesis is made available by the University of Alberta Library with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.