Defining the region of influence in traffic systems: methodology and implications
Restricted (Penn State Only)
- Author:
- Bai, Wushuang
- Graduate Program:
- Mechanical Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- November 25, 2024
- Committee Members:
- Robert Kunz, Professor in Charge/Director of Graduate Studies
Herschel Pangborn, Major Field Member
Satadru Dey, Major Field Member
Kshitij Jerath, Special Member
Vikash Gayah, Outside Unit & Field Member
Sean Brennan, Chair & Dissertation Advisor - Keywords:
- Connected and Autonomous Vehicles
Traffic Simulation
Traffic Dynamics
SUMO
Region of Influence - Abstract:
- Connected and Autonomous Vehicles (CAVs) are an emerging and highly impactful technology on today's roads. When assessing the performance of CAVs, it is useful to study their improvement relative to common metrics such as fuel economy/emissions, safety, and congestion. But metrics of an individual vehicle's performance alone may not be complete; for example, a CAV that is changing the signal timing on a smart traffic light to improve its own performance may degrade the performance of surrounding vehicles using the same intersection. Similar concerns arise in nearly all CAV topics: platooning, traffic light preemption, vehicle lane tracking, etc. Thus, assessment of a technology's impacts on surrounding traffic is important, possibly even more important than the improvements enabled on the CAV itself. While much research exists to measure and even control CAV behaviors even with surrounding traffic interactions, there is very poor definition of the formal meaning of "around the CAV". This leads to the question motivating this thesis: what boundary, associated with which metrics or factors, defines the vehicles, equipment, etc. "surrounding" a CAV? This work is motivated by a prior study investigating vehicles' performance in a local traffic community surrounding an ego vehicle. Using traffic simulators, the performance could be evaluated from low level to high level of vehicle automation and coordination. However, the selection of simulation domain was mostly based on users' experience and expertise. In the literature, one finds awareness of this concern when selecting the simulation domain, where instead of selecting the exact domain, the authors picked a wider area. A formal definition of simulation domains is still missing, and how to select a simulation domain remains unclear. Without a proper selection of simulation domain, it is difficult to assess whether the simulation completely encompasses all influences of factors under study. To address questions above, the primary goal of this work is to explore and develop methods to characterize the meaning of Region of Influence (ROI) for vehicle decision-making, where ROI is a formal definition relating to proximity in time and space to a vehicle and a behavior. This work first investigated ROI in a basic scenario: a vehicle's approach to a signalized traffic intersection. In this scenario, the ROI could be understood as the region over which knowledge of traffic light information is useful in informing fuel-efficient decisions on a vehicle's approach and traversal through the intersection. The author analyzed a heavy-duty vehicle's fuel consumption when driving through a green traffic signal as a baseline case, and the one when encountering a red traffic signal as a perturbed case. The ROI was studied using traffic simulation and then validated using field experiments on a test track. The results showed that first, the ROI of a traffic light begins approximately from 140 m to 200 m before an intersection and up to 175 m to 500 m after the intersection for heavy-duty diesel vehicles. Then this work extended the application of ROI into a more general scenario to include the ROI of a vehicle maneuvering on a highway. This problem was studied in an epsilon/delta formulation wherein a perturbation was exerted into a traffic system, and then simulations and theory were used to study how far in time and space the perturbation caused an influence on the surrounding system evolution. The results show that the ROI changes in time, changes with different perturbations, and can be different across different metrics. It was seen that highway traffic speed variance was found to have a strong influence on ROI evolution. In addition, this thesis characterized the time dimension of the ROI, specifically the initialization time for a traffic network to stabilize. The cumulative means of edge speeds on the entire network were studied and the 5% criterion was used to determine the settling time of a traffic simulation. This sets the foundation for an analysis of traffic performance in an even more general scenario, which is in an urban driving environment. It was found that the settling times for virtual and real-world networks are 0.75 and 0.82 hours. However, the ratio between the mean of settling times and the mean of trip completion times for the real-world network is higher than the one for the virtual network, approximately 20 times versus 10 times longer, respectively. Finally, this thesis studied the ROI of work zones, specifically lane closures in an urban traffic network. In this study, ROI was considered via a statistical analysis across both an idealized virtual grid network and a real-world network. The theory was validated against a number of traffic simulations with different random seeds to allow for stochasticity. The work zones were found to have an ROI that is approximately 2 - 3 times of the mean road network edge length for the virtual network; for the real-world network, the ROI is approximately 3 - 4 times of the mean network edge length.