Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Volume 615 Issue 7953, 23 March 2023

Driving test

One of the main hurdles to putting autonomous cars on the road is how to ensure the reliability of the artificial intelligence that replaces the human driver. Evaluating the safety of an AI driver to the level of a human in a naturalistic environment would require testing across hundreds of millions of miles — something that is clearly impractical. In this week’s issue, Henry Liu and his colleagues tackle this problem by training an AI to help test the AI in the driving seat. The researchers used dense deep reinforcement learning to train the AI tester, which allowed the tester to ignore safe scenarios and instead build a testing environment that focused on potentially dangerous situations. The team then successfully tested a real car using augmented reality — while the autonomous vehicle drove round a track it had to cope with virtual dangers set by the tester. The researchers say the system can speed up safety evaluations by several orders of magnitude.

Cover image: Shuang Li

This Week

Top of page ⤴

News in Focus

Top of page ⤴

Opinion

Top of page ⤴

Work

Top of page ⤴

Research

  • News & Views

  • Perspective

    • The current observations and understanding of black holes is reviewed, and the future of the field of black-hole astrophysics is discussed.

      • Ramesh Narayan
      • Eliot Quataert
      Perspective
  • Articles

Top of page ⤴

Amendments & Corrections

Top of page ⤴
Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing

Search

Quick links