The by-the-book approach is typically how the city’s recent fleet of self-driving vehicles are trained when they hit the roads, according to Carnegie Mellon University research professor Jeff Schneider. He was part of the team to found Uber’s autonomous vehicle division in 2015.
“We would never program our own cars to do something that, you know, wasn’t legally the right thing to do,” Schneider said. “The part that is not avoidable is dealing with the fact that other drivers on the road will be doing the Pittsburgh left.”
These vehicles are constantly collecting data about other drivers’ behaviors, which is where Schneider says machine learning comes into play.
“Essentially when we collect all the data from other cars driving around the streets, eventually the learning algorithms will see that pattern,” Schneider said.