快·开 (kuai kai)

Abstract

We propose a system for continuous evaluation of autonomous driving capabilities in high-speed racetracks. In the first implementation, the vehicles used on the racetrack will be small (1/16 or 1/10 of full-size). User-submitted controllers are evaluated remotely on real cars that can complete a racetrack circuit. Lap times are saved to a leaderboard that ranks submissions according to best (smallest) times. Potential users must pass simulation tests and, then, hardware-in-the-loop simulations, before being granted time with the real cars on the racetrack.