Automatically Testing Self-Driving Cars with Search-based Procedural Content Generation
Self-driving cars rely on software which needs to be thoroughly tested, but testing self-driving cars in real traffic is not only expensive but also dangerous, and has already caused fatalities. Virtual tests, in which self-driving car software is tested in computer simulations, offer a safer alternative, but creating suitable scenarios is laborious and difficult. In this paper we propose an automated technique that combines procedural content generation, a technique commonly employed in modern video games, and search-based testing in order to create challenging virtual scenarios for testing self-driving cars. Our AsFault prototype implements this approach to generate roads for testing lane keeping, one of the defining features of autonomous driving. Evaluation on two different self-driving car software systems demonstrates that AsFault can generate effective virtual road networks that succeed in revealing software failures, which manifest as cars departing their lane. Compared to random testing AsFault was not only more efficient, but also caused up to two times more lane departures.
Fri 19 Jul Times are displayed in time zone: (GMT+08:00) Beijing, Chongqing, Hong Kong, Urumqi change
|14:00 - 14:22|
|14:22 - 14:45|
Christian DegottCISPA Helmholtz Center for Information Security, Nataniel Borges Jr.CISPA Helmholtz Center for Information Security, Andreas ZellerCISPA Helmholtz Center for Information SecurityPre-print Media Attached
|14:45 - 15:07|
|15:07 - 15:30|