A team of researchers from the Virginia Tech Transportation Institute’s (VTTI) Division of Data and Analytics has developed an open-access interactive tool that seeks to enhance the capabilities of government, industry, and academic research in establishing automated driving systems (ADS).
The Operational Design Domain (ODD) Element Quantification Tool, funded by the National Institute of Standards and Technology (NIST), utilizes the institute’s cyber-physical systems framework that enables researchers and industry practitioners to understand and quantify automated driving operation conditions in a city, community, or region across and between 30 major U.S. cities.
However, data reflecting conditions from 30 cities is only the tip of the iceberg.
“This new tool does a large portion of the groundwork for organizations that are looking to expand their automated vehicles into new geographical areas,” said Gibran Ali, VTTI senior research associate. “The possibilities are endless as we have made it very easy to add in new territories or other datasets for analysis. This allows us to improve our capabilities in analyses of road networks, trip-taking behavior, ODD factors, and more.”
The ODD describes the conditions under which an automated vehicle is designed to operate. This includes its environmental conditions, roadway types, operational constraints, and geofence zones. Environmental conditions can be described through parameters such as temperature, humidity, and wind speed. Roadway and operational constraints are often provided in terms of functional class, number of lanes, and posted speed limits. This tool brings disparate sources of data into one place and helps automated vehicle developers design safer ADS-equipped vehicles.
For example, if a major automotive manufacturer wanted to understand how its vehicles would operate in Chicago, this prototype tool would allow them to both individually analyze the driving environment of Chicago and compare it to any other city in the platform, such as San Francisco or New York.
Although the driving environment contains substantial uncertainty, resources such as this tool can provide evidence-based descriptions of what an ADS-equipped vehicle may experience in its design domain. This project demonstrates how currently available datasets can be integrated and used to objectively measure the presence of key ODD elements, evaluate initial assumptions, and begin to establish baseline measurements that could be available to the larger development community.
“Working with VTTI on this project was a unique opportunity. VTTI has leading facilities and teams dedicated to assessing the safety of automotive systems,” said Edward R. Griffor, senior research scientist for cyber-physical systems and autonomous vehicles at NIST. “The ability to determine whether the capabilities of automated driving systems are sufficiently exercised in the testing environment is critical to understanding their safety, and this tool aims to show it is feasible.”
To assist in the development of the tool, among other data sources, VTTI incorporated fatal traffic crash data collected from the nationwide database called the Fatality Analysis Reporting Systems (FARS), which contains police-reported traffic crashes involving a motor vehicle traveling on a public roadway that resulted in the death within 30 days of the crash. Researchers analyzed data spanning 10 years and incorporated it into the tool’s framework. In addition to analyzing the FARS, census data was reviewed to better understand the populations across the 30 major U.S. cities that are currently available via the prototype tool.
“This portal serves as a starting point for conversations around how we can quantify the driving environment of automated vehicle driving,” said Mac McCall, VTTI senior research associate. “With the tool, researchers can change levels of individual factors such as speed limits and road types and immediately see the impact those changes have on the available road network and how those changes can impact their automated vehicles.”
The ODD Element Quantification Tool was developed by Ali, McCall, Kaye Sullivan, Shane McLaughlin, and Michelle Chaka.