How do you test whether a fully autonomous, self-driving car can cope with dangerous situations? In the real world accidents and near-misses are relatively rare — rare enough that it would take a lot of testing to see how a driverless car might react to different scenarios. Now a group of researchers has developed a new way to test autonomous vehicles, by surrounding them with virtual reality obstacles controlled by AIs. These virtual cars, trucks, deer, cyclists and pedestrians have been trained to recreate a variety of dangerous events based on real world data. The system is already being used to test and train autonomous vehicles currently in development.

Read the paper: https://www.nature.com/articles/s41586-023-05732-2

Subscribe to Nature Briefing, an unmissable daily round-up of science news, opinion and analysis free in your inbox every weekday.