On Learning the Tail Quantiles of Driving Behavior Distributions via Quantile Regression and Flows
This work addresses the need for accurate behavior models to improve autonomous driving simulations, but it is incremental as it adapts existing quantile learning frameworks to a specific domain.
The paper tackles the problem of learning models to capture the diversity and tail quantiles of human driver behavior distributions for safe autonomous driving, using quantile regression and autoregressive quantile flows on the highD dataset, and reports results in acceleration prediction and driver simulation rollouts with the tilted absolute loss metric.
Towards safe autonomous driving (AD), we consider the problem of learning models that accurately capture the diversity and tail quantiles of human driver behavior probability distributions, in interaction with an AD vehicle. Such models, which predict drivers' continuous actions from their states, are particularly relevant for closing the gap between AD agent simulations and reality. To this end, we adapt two flexible quantile learning frameworks for this setting that avoid strong distributional assumptions: (1) quantile regression (based on the titled absolute loss), and (2) autoregressive quantile flows (a version of normalizing flows). Training happens in a behavior cloning-fashion. We use the highD dataset consisting of driver trajectories on several highways. We evaluate our approach in a one-step acceleration prediction task, and in multi-step driver simulation rollouts. We report quantitative results using the tilted absolute loss as metric, give qualitative examples showing that realistic extremal behavior can be learned, and discuss the main insights.