The age of autonomous vehicles (AVs) is approaching, and the impacts they will have on transportation systems and policies will be profound. Transportation planners need to understand where AVs will emerge first, the patterns of AV trips, and the environmental impacts of AVs.
In a recent project, U of M researchers discovered that little is known about how characteristics of the built environment correlate with individuals’ acceptance of AV technologies.
“The few existing studies that looked at how the built environment could impact AV adoption used analysis areas that are too large for neighborhood-level planning,” says Tao Tao, a PhD candidate in the Humphrey School of Public Affairs. “This prohibited planners from dealing with the impacts of AVs in the communities where they will first emerge.”
To help transportation planners better understand where to focus their AV readiness efforts, Tao and his advisor, Professor Jason Cao of the Humphrey School, conducted research to determine how individuals’ interest in owning AVs in the future is associated with built environment variables and demographics.
“We identified three hypotheses that could be at work and wanted to see how important the three motivations were relative to one another,” Tao says.
The first hypothesis—innovation diffusion—says that people living in urban areas are among the first to accept new technologies. The efficiency hypothesis posits that people who live in suburban and rural areas will prefer to own AVs and enjoy their benefits. And the modal substitution hypothesis suggests that transit riders might treat AVs as a substitute for their transit trips.
Researchers used data from the regional travel survey (Twin Cities Travel Behavior Inventory) for the study. The working data was collected from October 2018 to September 2019 and included more than 6,000 households. In addition to collecting information on demographics, home locations, and travel, the survey asked respondents to indicate their interest in owning an AV in the future. Researchers analyzed the data using a machine-learning method known as a gradient-boosting decision tree, which has several advantages over traditional methods.
“This method is more effective in addressing irregular, nonlinear relationships than traditional statistical models,” Tao explains. “The gradient-boosting method combines hundreds of decision trees in a sequential order and makes them into a strong model.”
The results show that the innovation-diffusion hypothesis is the dominant motivation for owning AVs. Interest in owning AVs was positively associated with income and education and negatively associated with age.
The efficiency hypothesis was evident for three built environment variables. “In general, people who live in three types of areas—less mixed-use, lower population density, and with poorer road connections—show a higher interest in owning AVs,” Tao says. “They usually drive longer, so they value how AVs could help them make productive use of their travel time.”
The relationships between built environment variables and interest in owning AVs have implications for land use and transportation planning. Because AVs would let people use their travel time more effectively than conventional vehicles, they reduce the temporal cost of auto use—which, without policy interventions, would likely increase auto ownership.
“To prevent an increase in auto ownership due to AVs in urban areas, planners should work to enhance the attractiveness of travel mode alternatives,” Tao recommends. “In suburban and rural areas, growth management strategies could help mitigate potential sprawl. Additionally, strengthening express transit services to regional employment centers could be a promising strategy.”
This study was funded by a Smart and Connected Communities grant from the National Science Foundation.
Writer: Megan Tsai