IFML Seminar: 04/04/25 - Robust Autonomy Emerges from Self-Play

Abstract: Self-play has powered breakthroughs in two-player and multiplayer games. In this talk, I show that self-play is a surprisingly effective strategy in another domain: robust and naturalistic driving emerges entirely from self-play in simulation at unprecedented scale -- 1.6~billion~km of driving. This is enabled by Gigaflow, a batched simulator that can synthesize and train on 42 years of subjective driving experience per hour on a single 8-GPU node. The resulting policy achieves state-of-the-art performance on three independent autonomous driving benchmarks. The policy outperforms the prior state of the art when tested on recorded real-world scenarios, amidst human drivers, without ever seeing human data during training. The policy is realistic when assessed against human references and achieves unprecedented robustness, averaging 17.5 years of continuous driving between incidents in simulation.

 

Bio: Philipp is an Associate Professor in Computer Science at UT Austin, and leads a research team at Apple.