How human interactions with autonomous vehicles affect safety

By Meghan Chua


The idea of riding an autonomous shuttle around campus on the regular seems closer to reality after the Wisconsin Automated Vehicle Proving Grounds shuttle demonstrations last week. But before autonomous vehicles can become mainstream, the people who will commute in them have to be comfortable with the technology.

Graduate student Hannah Silber
Graduate student Hannah Silber studies how humans interact with autonomous vehicle technology.

For graduate student Hannah Silber, that entails closing the gap between an autonomous vehicle’s “brain” and the human one.

Silber, a PhD candidate in Civil and Environmental Engineering who plans to graduate in May, studies human factors and safety in transportation and aviation. She spoke about her research on human-automation interactions Thursday at a demonstration of the WiscAV Autonom Shuttle.

In one project, Silber explored how a driver who is engaged in talking to a passenger or listening to music – both auditory tasks – differs from a driver engaged in another visual task, such as changing the settings on a Bluetooth dashboard in the vehicle. The second situation shows heightened workload on the driver, putting multiple demands on the driver’s vision.

That makes a case for autonomous vehicles being safer. However, that doesn’t mean they’re perfect.

“You might see that safety is improved because the autonomous vehicle isn’t necessarily distracted by the things going on while it’s driving, but its brain isn’t flexible like the human brain, which is adaptable to a lot of different situations,” Silber said.

The WiscAV Autonom Shuttle
People line up to get a look at the WiscAV Autonom Shuttle during an on-campus demonstration April 25.

In a busy four-way intersection on campus, for example, a human driver knows pedestrians will cross in front of them. The driver will inch forward to show pedestrians they are going to move, and eventually get through the intersection. An autonomous vehicle, however, will sense that there are people or vehicles moving in the intersection and won’t move at all, possibly getting stuck for a while.

“There’s a gap between how [the autonomous vehicle] will behave on the roadway and how human drivers behave,” Silber said.

Streamlining the algorithms that make autonomous vehicle “brains” think more like humans can help close that gap, she said. As the autonomous vehicle industry continues to address these questions, researchers at UW–Madison will be among those seeking to answer it.

“What I’ve found here is that everyone is always trying to move forward with research,” Silber said. “It’s really about: this is what’s happening in the research world now, so how can we be a part of this?”