Developing Improved Models for Assessing the Impacts of Automation on Transportation Operations
Date and Time: Tuesday, July 11, 2023: 5:30 PM - 7:00 PM
Presentation Description
Connected and automated vehicles (CAV) are a promising technology that has the potential to revolutionize the transportation industry by providing several benefits, including increased safety, reduced congestion, and improved efficiency. The technology has received significant investments from both private industry and public agencies in recent years. Different levels of vehicle automation systems, such as advanced driver assistance systems (ADAS), have already started to travel on public roads. The impacts of automation on transportation operations must be accurately assessed to support different transportation agencies in preparing for the deployment and adoption of CAV technology. While several studies have been conducted on the impacts of CAV technology on transportation system performance, most of these studies were not calibrated and verified using ADAS-equipped vehicle field collected data to accurately reflect actual ADAS driving behaviors or the behavior of human driven vehicles while interacting with ADAS-equipped vehicles. This research makes two key contributions. First, this research developed an improved ADAS-equipped vehicle car-following model capable of capturing the nonlinearity of adaptive cruise control (ACC) applications. Additionally, this research developed an improved nonautomated vehicle (SAE L0) car-following model that captures how human drivers alter their car-following behavior when interacting with both conspicuous and inconspicuous ADAS-equipped vehicles. Both models are developed and validated using data collected with ADAS-equipped vehicles in live, naturalistic traffic.
This study utilizes field collected datasets from two Federal Highway Administration (FHWA) data collection projects: the Ohio and Third Generation Simulation (TGSIM) datasets. To develop the nonlinear ACC model, the project team used data collected using two commercially available SAE level 2 ADAS-equipped vehicles in central Ohio to develop a multivariate piecewise linear adaptive cruise control model to investigate longitudinal behaviors of ACC. Five days of collected data were used in model development, with 70 percent of data used for model calibration and 30 percent of data used for model validation. According to the results, the developed model captured ADAS ACC behaviors well, without overfitting to the calibration data. Future work will utilize the nonlinear ACC car-following model in case studies to explore the impact of ACC on transportation system operations.
Concurrently, the project team is using the TGSIM datasets to develop an improved human driven vehicle car-following model that can capture how nonautomated vehicle (i.e., completely human driven vehicles) alter their behavior in the presence of ADAS-equipped vehicles. Car-following models are one of the fundamental models to emulate driving behaviors in microscopic, traffic operation models. To study the impacts of ADAS-equipped vehicles on the driving behaviors of human driven vehicles, the project team is calibrating the Intelligent Driver Model (IDM) using the TGSIM datasets. The TGSIM dataset was collected using multiple commercially available SAE level 1 and level 2 ADAS-equipped vehicles. The TGSIM project also used helicopters to capture vehicle trajectories of an entire highway segment. As a result, TGSIM datasets are ideal for investigating interactions between humans and ADAS-equipped vehicles under diverse scenarios in freeway environments. Specifically, this study is using a machine learning-based methodology to calibrate the IDM model. This model is currently under development, and we expect to have preliminary results in time for ARTS 2023.
RESULTS:
The developed multivariate piecewise linear adaptive cruise control model can properly emulate SAE L1/L2 ADAS-equipped vehicle behaviors when their ACC functions are activated. Additionally, the calibrated car-following model expects to more accurately capture changes in the driving behaviors of nonautomated vehicle (SAE L0) when they interact with CAVs. These models will help state and local transportation agencies to propose various strategies and policies to better prepare for and facilitate the development of CAV technology, especially at the early stage of deployment of CAV technology.
Speaker Biography
Dr. Rachel James recently joined the Federal Highway Administration Office of Transportation Policy Studies within the Office of Policy and Governmental Affairs as a Policy Analyst in May 2023. Prior to this assignment, she was the FHWA Connected and Automated Vehicle Analysis, Modeling, and Simulation Research Program Manager in the FHWA Office of Safety and Operations Research and Development. Rachel’s career goal is to help bridge the gap between transportation research and policy. She is passionate about using data to develop improved analysis, modeling, and simulation tools to inform better operational and investment decisions. Dr. James received her B.S. in Civil Engineering from West Virginia University in 2014 and her M.S. and Ph.D. in Civil Engineering (Transportation) from The University of Texas at Austin in 2016 and 2019, respectively.
Presentation File
Developing Improved Models for Assessing the Impacts of Automation on Transportation Operations
Category
Poster
Description