Research that began during Kettering University's participation in the SAE AutoDrive Challenge has been published in the peer-reviewed journal AI & Society and featured in Bioengineer.org.
Conducted by Kettering students Olivia Wanless ‘20, M'22 (now Olivia Gettel), Chanc Gettel '21 and Charlie Gates M'20, the study identifies gaps between where autonomous vehicles (AVs) are tested and the communities and conditions they may eventually encounter.
"We came up with the concept on the way to an AutoDrive Challenge event in Pittsburgh," said Olivia Gettel. "As part of the challenge, we participated in 'Social Responsibility' events that required us to think about the social impact of autonomous vehicles."
For Olivia and Chanc, who grew up in a rural area, the question was particularly relevant. Limited charging infrastructure and difficult winter road conditions were among the concerns surrounding the technology in their community.
"The general perception of AVs from our area seemed to be that they wouldn't be very useful," Gettel said. "When we started researching that for this paper, it really drove home that a lot of those concerns were valid."
Dr. Diane Peters, Professor of Mechanical Engineering and leader of Kettering's AutoDrive program, and Dr. James Huggins, Associate Professor of Computer Science, collaborated with the students on the research and subsequent publication.
"These students could see that this was something important that needed to be addressed," Peters said. "Jim and I contributed, but they were the ones who set the direction, did the majority of the work and shaped the key aspects of the paper."
Where We Test Matters
Of the 33 U.S. cities documented as having begun or committed to autonomous vehicle testing in 2019, none qualified as rural under the study's population-density definition. The researchers later incorporated updated demographic data and six additional testing cities documented in 2025. Urban locations continued to dominate the expanded dataset.
Rural roads present conditions less common in metropolitan testing environments, including gravel surfaces, agricultural equipment, wildlife and missing lane markings. Snow can create additional challenges when the roadway itself becomes difficult to distinguish.
"In a rural area, it's not uncommon to be the first vehicle to forge a path on an unplowed road. It can be challenging to see where the roads actually are," Gettel said. "If AVs aren't being tested in those areas, they won't be safe to operate in those conditions."
The researchers also examined demographic and community characteristics of testing locations, including race and ethnicity, disability prevalence, gender, bicycling activity and economic conditions. Their analysis raises questions about whether documented testing locations adequately reflect the diversity of pedestrians, cyclists and people with different mobility needs that autonomous vehicles may encounter.
"Where we test matters," Huggins said. "Generally speaking, it is easy for technology developers to simply test their products in whatever environments are most convenient. But if the dream of autonomous vehicle development is to be achieved, then autonomous vehicles have to work in every environment, not just convenient environments."
To address those gaps, the researchers propose more representative testing criteria, including conducting 30% of testing miles on rural roads, reflecting the approximate national distribution of vehicle miles traveled. They also recommend accounting for different populations and transportation behaviors, along with documenting interactions between AVs and road users.
"Just as we have different drive cycles for urban areas versus highways versus rural roads, we should have different test criteria that are aimed at all of the different places that we'll be operating autonomous vehicles, and ensure that they are inclusive of everyone," Peters said.
