Robust autonomy emerges from self-play
Self-play has powered breakthroughs in two-player and multi-player games. Here we show that self-play is a surprisingly effective strategy in another domain. We show that 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.
GIGAFLOW models uncertainty directly through noise on the state S
What king of noise? uniform? pink? brown? This made me curious because couldn't uncertainty me modeled by a distribution obtained from real-data observations?
GIGAFLOW agents observe the positions and speeds of nearby agents but not their acceleration
Why? Humans perceive (above a minimum threshold time) if another vehicle is accelarating. Couldn't the agents be constructed in a way for them to construct an estimate of accelaration based on previous N observations other veihcles?