Embedding Synthetic Off-Policy Experience for Autonomous Driving via Zero-Shot Curricula

TL;DR

This paper introduces a method to predict the difficulty of driving scenarios and uses this to create curricula for training autonomous driving agents, improving safety and efficiency with less data.

Contribution

It presents a novel difficulty prediction approach and demonstrates zero-shot curriculum transfer that enhances autonomous driving performance with reduced data.

Findings

Reducing training data to 10% maintains performance levels.

Prioritizing difficult scenarios reduces collisions by 15%.

Increases route adherence by 14%.

Abstract

ML-based motion planning is a promising approach to produce agents that exhibit complex behaviors, and automatically adapt to novel environments. In the context of autonomous driving, it is common to treat all available training data equally. However, this approach produces agents that do not perform robustly in safety-critical settings, an issue that cannot be addressed by simply adding more data to the training set - we show that an agent trained using only a 10% subset of the data performs just as well as an agent trained on the entire dataset. We present a method to predict the inherent difficulty of a driving situation given data collected from a fleet of autonomous vehicles deployed on public roads. We then demonstrate that this difficulty score can be used in a zero-shot transfer to generate curricula for an imitation-learning based planning agent. Compared to training on the entire unbiased training dataset, we show that prioritizing difficult driving scenarios both reduces collisions by 15% and increases route adherence by 14% in closed-loop evaluation, all while using only 10% of the training data.