Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


37958.pdf (8.66 MB)

ETD Abstract Container

Abstract Header

Connected and Automated Traffic Control at Signalized Intersections under Mixed-autonomy Environments

Guo, Yi
http://rave.ohiolink.edu/etdc/view?acc_num=ucin1613752599541812

Abstract Details

2020, PhD, University of Cincinnati, Engineering and Applied Science: Civil Engineering.
Connected and automated (CAV) technologies offer new opportunities that can transmit real-time vehicular information, and their trajectories can be precisely controlled, which eliminates the barriers of conventional control framework. However, before the CAVs are prevalent in the traffic stream, a mixed-autonomy environment will last a long period with gradually increasing CAV market penetration. This indicates that traffic management under mixed-autonomy environment is essential in the transition from conventional traffic management to full-autonomy traffic control. To signalized intersection, vehicular trajectory control and signal optimization based on CAV technologies are two approaches that have significant potential to mitigate congestion, lessen the risk of crashes, reduce fuel consumption, and decrease emissions at intersections. Therefore, these two approaches should be integrated into a unified traffic management framework such that both aspects can be optimized simultaneously to achieve maximum benefits. This dissertation proposes a mixed-autonomy traffic management framework to integrate the signal control and trajectory control systematically. The framework architecture consists of six layers, including sensing, information, planning, optimization, control, and evaluation, and each layer has its own scope and responsibility. The proposed framework is a flexible and compatible framework for joint optimization of vehicle trajectory and signal control, and it can be applied for both full-autonomy and mixed-autonomy environments. The development details of major components are also described. A dynamic programming (DP) framework with trajectory planning with piecewise polynomials (TP3) as a subroutine (DP-TP3) is presented to solve the joint optimization of signal control and vehicle trajectory control considering conflicts of the four movements. The proposed TP3 algorithm provides an analytically solvable operation for vehicular trajectory construction and prediction, and the interactions between human-driven vehicles (HDVs) and CAVs are explicitly considered. Each phase in a cycle is regarded as a stage, and the DP algorithm considers both traffic system performance and individual performance at each stage to obtain an optimal signal plan. After an optimal signal timing solution is obtained through DP-TP3, a customized numerical gradient-based approach will be run to optimize for best system performance in terms of travel time, fuel consumption, and safety. Numerical experiments are conducted to investigate the performance of DP-TP3, and results show that the average travel time can be reduced by up to 35.72%, and the fuel consumption can be reduced by up to 31.5%, compared with the adaptive signal control. To implement an efficient learning control, the DP is replaced with deep reinforcement learning (DRL), and a deep reinforcement learning framework with trajectory planning with piecewise polynomials (DRL-TP3) is proposed. The signal controller is trained by minimizing the differences between the estimated values and actual values with iterative interactions between the signal controller and traffic environment, and queue length and delay are combined as the reward. A long short-term memory-based HDV state estimation is developed to estimate HDV states for supporting trajectory planning and signal optimization. Microscopic simulations are conducted, and the impacts of DRL-TP3 and subset strategies across different market penetration rates are simulated and discussed.
Jiaqi Ma, Ph.D. (Committee Chair)
Na Chen, Ph.D. (Committee Member)
Kelly Cohen, Ph.D. (Committee Member)
Nabil Nassif, Ph.D. (Committee Member)
131 p.
Civil Engineering
Connected and Automated Vehicle; Trajectory Control; Traffic Signal Control; Dynamic Programming; Deep Reinforcement Learning

Recommended Citations

Citations

  • Guo, Y. (2020). Connected and Automated Traffic Control at Signalized Intersections under Mixed-autonomy Environments [Doctoral dissertation, University of Cincinnati]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1613752599541812

    APA Style (7th edition)

  • Guo, Yi. Connected and Automated Traffic Control at Signalized Intersections under Mixed-autonomy Environments. 2020. University of Cincinnati, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=ucin1613752599541812.

    MLA Style (8th edition)

  • Guo, Yi. "Connected and Automated Traffic Control at Signalized Intersections under Mixed-autonomy Environments." Doctoral dissertation, University of Cincinnati, 2020. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1613752599541812

    Chicago Manual of Style (17th edition)

ucin1613752599541812
376
© 2020, all rights reserved.
This open access ETD is published by University of Cincinnati and OhioLINK.
Release 3.2.12