In Pittsburgh, Ding Zhao, an assistant professor of mechanical engineering at Carnegie Mellon, has been working with carmakers to use digital twins to improve the safety of self-driving vehicles. In his lab, he leverages vast quantities of data collected from real tests of self-driving cars to build complex digital-twin simulators. The simulations, he says, help predict how a car’s AI will react in dicey situations that could be dangerous and difficult to re-create IRL: when merging onto a dark snowy highway, for instance, or when jammed in between two trucks.
Crucially, digital twins also allow researchers to run crash-test simulations countless times without having to destroy cars or endanger real people. That means digital-twin technology is becoming essential to the development of self-driving cars. “Real-world testing is too expensive and sometimes not even effective,” Zhao says. Digital twins are also being used in other complex and potentially dangerous machines, from nuclear reactors in Idaho to wind turbines in Paris.