For more than 30 years, Nintendo’s Mario Kart has been one of the world’s most popular racing games. However, safe driving in the game doesn’t necessarily mean winning the race. So, can this classic game be reengineered to help autonomous driving systems learn how to drive safely? Researchers at the University of Maryland believe it can.
They are putting the joystick and controller in the virtual hands of autonomous driving programs before these systems are tested on real roads.
“We’re actually using these automated driving systems to figure out how to verify that the systems are correct, and then make sure that they adhere to the rules of the road, and design ways that we can make these better if they don’t, if they violate these rules,” said Mumu Xu, associate professor of aerospace engineering.
The approach is simple: handle Mario Kart safely, and then build from there.
“We train Mario to drive around a lap,” Xu explained in an interview with WTOP. “We train him to go fast and go safe. Based off of those traces that he drives, we take those and we can give you a score to say how safe you are.”
Using that score, the researchers can provide feedback to designers to improve the safety of the autonomous driving systems.
“We can say to the designers, go back and train your car again to make sure that it doesn’t violate certain rules, it speeds, or it bumps into other cars,” Xu added.
The program is still in its very early stages, but Mario happens to be a popular figure in the AI world.
“Using Mario is a way for us to work in simulation, to first test whether or not something is useful,” Xu said. “Then, we can go out onto the roads and then make sure that things are actually correct.”
—
Get breaking news and daily headlines delivered to your email inbox by signing up [here](#).
© 2025 WTOP. This website is not intended for users located within the European Economic Area.
https://wtop.com/maryland/2025/10/can-a-video-game-help-autonomous-driving-programs-learn-how-to-drive/
