Uber crash shows need for more simulated testing of autonomous cars, University at Buffalo computer scientist says

BUFFALO, N.Y. — The fatality involving a pedestrian in Arizona struck by an autonomous vehicle operated by Uber illustrates the need for more rigorous simulated testing of driverless cars, says University at Buffalo computer scientist Chunming Qiao.

Qiao, who studies the computing systems behind driverless and connected cars, stressed that road-testing of autonomous vehicles is necessary and valuable to advancing self-driving technology that he thinks will ultimately save lives and make roadways safer.

But he says that companies should be doing more simulated testing before rushing into extensive road testing.

“Right now, the technology is so new. With all these companies doing road testing, there will be accidents. It’s not a question of if. It’s a question of when,” says Qiao, the lead investigator of a National Science Foundation-funded research project called iCAVE2, which stresses simulated testing of connected and self-driving vehicles.

“To win consumer confidence and market acceptance, non-biased third parties should be defining standard testing scenarios that each company must use. Those parties would then analyze the performance of each company and issue safety certificates before self-driving cars can be commercialized.”

Because road testing is costly and time-consuming, a more practical approach is to use high-fidelity, virtual reality-based simulators,” Qiao says.

“These simulators, including the one we are building, look like large arcade games. We can simulate a driverless car with an endless number of driving scenarios, and under various hardware or software failures, which the car then reacts to. These scenarios are also repeatable, which means you can run multiple tests to ensure the technology’s safety. You simply cannot do that on the road,” he says.