Technology is revolutionizing the way we drive. Many technological innovations designed to increase driver safety are already available, and many more are being developed. Some of these safety systems are designed to help drivers make appropriate decisions, while others will initiate decisions themselves if the driver is not capable during safety-critical moments. However, these systems can come with unanticipated effects—causing the driver’s behavior to change or adapt in unforeseen ways that may compromise the potential benefits of a system.
Researcher Linda Ng Boyle aims to make sure the adaptive effects of vehicle safety systems are fully understood. As professor and chair of the Department of Industrial and Systems Engineering at the University of Washington, she has conducted numerous studies that model the effects of drivers’ adaptive behavior on system safety. In October, Boyle visited the U of M to deliver a presentation on her work as part of the Roadway Safety Institute’s Seminar Series.
According to Boyle, behavioral adaptation can have a significant effect on the performance of driver safety systems.
“The idea behind behavioral adaptation is that when you use a driving technology, your performance will initially improve, but over time your performance will level off,” Boyle said. “If the system is deactivated, there may be a positive transfer, where what we learn remains with us to make our performance better than before, but the more likely case is when the system is gone you’ll go back to your baseline performance. There’s also the chance of a negative transfer where your performance actually gets worse when the system is deactivated.”
One specific technology that is influenced by driver adaptation is adaptive cruise control (ACC), which automatically adjusts the vehicle speed to maintain a safe distance from the vehicles ahead. In her presentation, Boyle examined how ACC can be used as a case study to show how behavioral adaptation can be studied and modeled using two different approaches: naturalistic studies and driving simulation studies.
“In naturalistic studies that use real-world data, there can be a lot of variation in the data, but information and observations from the real-world environment can show us what particular situations to simulate in the controlled environment of a driving simulator,” she said.
During the past five years, Boyle has conducted several studies to help shed light on the adaptive effects of ACC. For example, a naturalistic study analyzed real-world data to determine the likelihood of drivers manually braking when the ACC system began braking or slowing down the vehicle. A second study used a driving simulator to test how experienced ACC drivers used the driving technology, how often they engaged or disengaged the ACC, how many warnings the ACC system issued, and how much drivers trusted their ACC system.
“We found that though the group of drivers was pretty homogenous in terms of age, they all used ACC very differently,” Boyle said. “Based on these findings, we were able to classify drivers into three different cluster groups—conservative, moderately risky, and risky.”
Boyle believes that the best way to effectively model drivers’ adaptive behavior is to use a balanced approach that takes into account both naturalistic and simulation studies. “You need a range of different techniques and an understanding of how to bring them together,” Boyle said. “In this way, we’ll be able to learn how to design better systems that can reduce the risk and enhance the safety of everyone using the transportation network.”
A recording of the seminar is available on the Roadway Safety Institute website.