ITS Berkeley Awards 7 STRP Projects
In August, 2024 the UC Berkeley Institute of Transportation Studies (ITS) awarded seven new research proposals, totaling $595,000 through the Statewide Transportation Research Program (STRP) funded by the State of California through the Road Repair and Accountability Act of 2017 (Senate Bill 1). Two of the proposals were awarded to SafeTREC Director Julia Griswold:
Autonomous Vehicle Safety Performance in Mixed Traffic: Insights from NHTSA Crash Data
PI: Julia Griswold
Project Description:
As Autonomous Vehicles of Level 3 or higher gain prominence, numerous manufacturers are testing these vehicles on U.S. public roads, particularly in California. Consequently, the safety performance of these AVs has emerged as a critical issue within the transportation industry. When AVs are involved in a crash, the National Highway Traffic Safety Administration (NHTSA) and the California Department of Motor Vehicles (DMV) have mandated that AV companies report these crashes under a General Order. These AV crash reports provide detailed descriptions. However, a clean and preprocessed dataset for AV crashes is not available to provide insights into safety performance and real-world crashes.
This proposed project will enhance accessibility to AV safety-related data by extracting useful information from AV crash narratives from the NHTSA dataset, using Natural Language Processing techniques and a manual review of the narratives, as well as preprocessing the data to create a unique real-world AV crash dataset. This dataset can help safety practitioners, researchers, and stakeholders to actively contribute to building a collective understanding of AV safety. Furthermore, by analyzing the dataset, we will cluster real-world AV crash scenarios using an unsupervised machine learning technique to identify rare (strong edge case), complex (weak edge case), and usual scenarios in real-world AV crashes. Furthermore, we will identify the contributing factors of each class and focus particularly on the human behavior of those involved in AV crashes (i.e., crash partners). This can aid developers in enhancing AV technologies and tackling safety concerns in mixed traffic, which will help better manage public expectations and alleviate potential apprehensions.
A Time and Space Exploration of Traffic Crash Trends During the COVID Recovery
PI: Julia Griswold
Project Description:
The proposed research aims to leverage post-pandemic congestion patterns to understand how reduced congestion impacts traffic crashes. Utilizing freely available data from Caltrans' PeMS platform and crash records from SafeTREC's TIMS platform (SWITRS data), this study will incorporate congestion metrics into the predictive method used for network monitoring and infrastructure design, prescribed in the Highway Safety Manual (HSM) (2). Currently, the predictive method employs generalized linear regression models (e.g., Poisson or Negative Binomial) to model crash occurrences as a function of Average Daily Traffic (ADT), without accounting for whether traffic occurs in free flow or congested conditions.
The project will retrieve data from the aforementioned platforms to construct a dataset of highway segments with traffic count-derived data, Vehicle Miles Travelled (VMT), Vehicle Hours Travelled (VHT), and traffic crash counts for various severity levels recorded in the SWITRS data. This research will ultimately enhance our understanding of how congestion affects safety performance, complementing current methodologies.
The compiled data will be used to calibrate and test predictive models similar to those prescribed by the HSM, incorporating congestion metrics derived from VHT data. The calibration process will identify when the new variables are significant, while the testing phase will ensure models do not suffer from overfitting. Additionally, the project will conduct a series of tests to identify locations (from those compiled in the dataset) that would be detected as High Crash Concentration Locations using both the traditional and proposed methods. The results will highlight the merits of using congestion as an explanatory variable for predicting traffic crashes.
Additional 2024 ITS Award recipients include:
- Michael Cassidy: Congestion Pricing to Support Transit
- Ken Alex: An Open-Access Tool for Equity-Focused Local Electric Mobility Investment Planning
- Daniel Chatman: Understanding the Impacts of Working at Home and Online Shopping on Post-Pandemic Travel and Transportation Policy in California
- Marta Gonzalez: The Interplay of Remote Work, Economic Complexity, and City Structure in Reshaping Mobility Dynamics at the Individual Level and a Metropolitan Scale
- Joshua Meng: Bridging Transit Service Assessment and Community Needs for Equitable Mobility: A Case Study in San Francisco Chinatown
Visit the ITS website for more information about the 2024 ITS Award recipients for SRTP Projects.