There’s a fine balance between safety and innovation.
That’s certainly the case with autonomous vehicles, but also with many pioneering technologies. How do you improve the former without compromising the latter?
Professor Huei Peng, Director of Mcity, believes that innovation in autonomous vehicles will solve many safety issues that drivers face on the road today.
But there’s also the problem of public perception. Despite the thousands of driver-responsible car accidents that occur in the United States every day, it’s the odd accident involving an autonomous vehicle that makes the news.
Peng and his colleagues at Mcity have developed the ABC Test, a three-tiered strategy that assesses autonomous vehicle safety in a controlled environment before letting the vehicles out on public roads. In doing so, the ABC Test aims to help developers fine tune their autonomous vehicles while also establishing a demonstrated track record of success.
“People have this idea that autonomous vehicles are driving under any circumstance,” Peng says. “But look at early planes. You couldn’t fly high. You couldn’t fly under bad weather. It’s the same idea.”
Assessing safety performance on an AV test track
The ABC Test is a three-tiered lab assessment of an autonomous vehicle. It just so happens that this laboratory is a test track.
The A in ABC stands for Accelerated Evaluation, the B for Behavior Competence, and the C for Corner Cases.
In the accelerated evaluation portion of the test, researchers identified riskier yet still-common driving actions. Rather than focus on minimum risk scenarios, researchers focus on maneuvers like others cutting in front of the car or making left turns.
“The idea is that these common scenarios deserve increased attention,” Peng says.
In behavior competence, researchers will submit vehicles to a list of 50 of the most common-yet-risky driving scenarios to demonstrate safety performance.
And for corner cases, researchers then take those driving scenarios and complicate them further by challenging vehicles to perform under more demanding situations like low-light conditions or having to contend with bicycles and motorcycles weaving in and out of traffic. Even the color of pedestrians’ clothing is a variable.
Mcity is currently focusing on 16 of the behavior competence scenarios and will continue to do so through August of this year. They are waiting to learn more from those tests before executing the rest of the 50.
“We need to make sure we execute these tests and show people that these are the procedures, and that we hold high standards,” Peng says.
Addressing the 40 percent
While the government has mentioned safety performance testing, there is no mandate to do so. Mcity, along with its partners in the automotive industry, is taking it upon itself to come up with a model for effective safety assessment.
That’s just fine with Peng. He believes that it should remain in the hands of research and development.
There are a lot of safety problems that autonomous vehicles can address. Distracted driving. Aging drivers. And despite some people’s fears that autonomous vehicles will take away jobs, the United States is facing a truck driver shortage. It’s a difficult job and perhaps autonomous vehicles can fill those employment gaps.
Still, there’s a lot of work to be done to convince the general public that autonomous vehicles are safe. As referenced in Peng’s white paper
Mcity ABC Test: A Concept to Assess the Safety Performance of Highly Automated Vehicles, an August 2018 survey by J.D. Power found that 40 percent of the respondents “would not ride in an (automated vehicle) regardless of what progress is made.”
And that’s why they’ve developed the ABC Test, says Peng. “That’s why we’re working on best practices.”
Photos courtesy of University of Michigan