Infrastructure Preparation for Connected and Automated Vehicle Deployment
Restricted (Penn State Only)
- Author:
- Chen, Chenxi
- Graduate Program:
- Civil Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- September 28, 2024
- Committee Members:
- Xianbiao Hu, Chair & Dissertation Advisor
S. Ilgin Guler, Major Field Member
Zhitong Huang, Special Member
Vikash Gayah, Major Field Member
Guohong Cao, Outside Unit & Field Member
Jay Regan, Professor in Charge/Director of Graduate Studies - Keywords:
- Connected and Automated Vehicles
Infrastructure
User equilibrium - Abstract:
- Connected and Automated Vehicle (CAV) is poised to transform transportation by reducing accidents, emissions, and congestion. However, widespread implementation faces numerous significant challenges, with infrastructure preparation emerging as a particularly critical and complex concern. Previous research has consistently shown that both physical and digital infrastructure will be significantly affected by CAVs and require substantial modifications to support their deployment. The primary objective of this dissertation is to explore how infrastructure can be effectively prepared for the deployment of CAV. Firstly, the distinctive driving patterns of CAV demonstrated to affect various aspects of physical infrastructure, from geometric design to pavement performance. For example, the more concentrated lateral wandering distribution of CAV, resulting from lane-centering technology, is likely to accelerate pavement deterioration. Therefore, it is essential to ensure that the physical transportation infrastructure is prepared to accommodate this unique driving pattern of CAV. Secondly, the development and deployment of digital infrastructure systems, such as road-side sensors and communication devises, are crucial for the development of CAV technology to achieve a safe and efficient operation. At an early stage of CAV deployment, mixed traffic flow with CAVs and human-driven vehicles (HDVs) will exist for a long period. Preparation of digital transportation infrastructure will bring benefits for both CAVs and HDVs without high-performance sensors. Furthermore, as both physical and digital transportation infrastructures become more sophisticated, accurate modeling of the interactions between CAV traffic and these infrastructures is essential. For instance, understanding the user equilibrium pattern of CAVs traffic at bottlenecks can guide us in designing systems that optimize traffic flow, and reduce energy consumption. Studies have been initiated to investigate the potential impact of CAV on transportation infrastructure to support future CAV testing and deployment. However, most existing research only focuses on the different wandering patterns of CAV. To bridge this gap, an apple-to-apple comparison is first performed to systematically reveal the behavioral differences between the human-driven vehicle (HDV) and CAV trajectory patterns for the first time, with the data collected from camera-based Next Generation Simulation (NGSIM) dataset and autonomous driving co-simulation platform, CARLA and SUMO, respectively. A gradient boosting-based ensemble learning model for pavement performance (i.e., International Roughness Index, IRI) prediction is then developed with the input features including three driving pattern features, namely, lateral wandering deviation, longitudinal car-following distance, and driving speed, plus other twenty context variables. A total of 1,707 observations is extracted from the Long-Term Pavement Performance (LTPP) database for model training purposes. The result indicates that the trained model can accurately predict pavement deterioration, and that CAV deteriorates pavement faster than HDV by 8.1% on average. The results of sensitivity analysis show that CAV deployment will create a greater impact on the younger pavements and the pavement performance will be affected more significantly with a lower penetration rate. In addition, the rate of pavement deterioration is found to be stable under light traffic, whereas it will increase under congested traffic. Beyond preparing physical infrastructure, the implementation of CAV technology necessitates an array of sensors to ensure safe and efficient operation. Sensors embedded in infrastructure hold significant promise for early-stage adoption of CAV technologies, despite the relatively high costs associated with advanced sensors like LiDAR and GNSS, which pose barriers to widespread deployment. The advantages of digital infrastructure manifest in two key ways: for connected vehicles with a lower level of automation lacking onboard perception sensors, infrastructure-based sensors substantially enhance their contextual awareness. For highly automated vehicles equipped with comprehensive sensor suites, these infrastructure sensors help mitigate issues related to occlusion and limited sensor range. This manuscript introduces a cooperative perception modeling framework that addresses a critical technical challenge: the time delay in cooperative perception, crucial for synchronization, perception, and localization processes. We first develop a Constant Turn-Rate Velocity (CTRV) model to predict vehicle motion states. Subsequently, we introduce a delay compensation and fusion module to address time delays caused by processing and communication latencies. Furthermore, given the non-linear nature of the movements of vehicles, cyclists, and pedestrians in terms of position and speed. We implement an Unscented Kalman Filter (UKF) algorithm to enhance the accuracy of object tracking, factoring in the communication delays between the vehicle and infrastructure-based LiDAR sensors. Simulation tests are conducted to assess the feasibility and effectiveness of the proposed algorithm, yielding encouraging results. With the advancement of both physical and digital infrastructure, modeling the dynamic and interactive relationship between traffic demand and supply is crucial. However, existing studies often neglect the impact of the dynamic effective discharge rate on user equilibrium, making it challenging to predict overall traffic flow for effective traffic management. To deal with this, this manuscript proposes a demand-supply coupled equilibrium model for the morning commute problem, incorporating the dynamic effective discharge rate into the traffic model. First, this study analytically derives the reduced discharge rate and merging delay for both mainline and ramp traffic. Considering the effective discharge rate and merging delay. We analyze and solve the user equilibrium profile in a generic Y-shaped network under different scenarios. The time-varying demand-supply relationship determines the network congestion pattern, leading to varying possible scenarios. The model is validated in a typical scenario with bottlenecks at the upstream, downstream, and the merging area. We analyze and compare the departure time profile and costs under user equilibrium with results from the classic model that ignores the effective discharge rate and merging delay. Our findings indicate that the classic model underestimates the cost for commuters affected by the reduced discharge rate and delay in the merging area, necessitating an earlier departure time to achieve user equilibrium.