What Matters for Scalable and Robust Learning in End-to-End Driving Planners?

TL;DR

This paper critically examines architectural patterns in end-to-end driving models, revealing limitations and synergies, and introduces BevAD, a scalable architecture that achieves high success rates and strong data scaling in closed-loop driving tasks.

Contribution

It systematically analyzes the impact of perceptual, trajectory, and generative planning patterns on closed-loop performance and proposes BevAD, a new scalable architecture for autonomous driving.

Findings

BevAD achieves 72.7% success on Bench2Drive.

Architectural patterns have complex, sometimes unexpected effects on closed-loop performance.

BevAD demonstrates strong data-scaling behavior with imitation learning.

Abstract

End-to-end autonomous driving has gained significant attention for its potential to learn robust behavior in interactive scenarios and scale with data. Popular architectures often build on separate modules for perception and planning connected through latent representations, such as bird's eye view feature grids, to maintain end-to-end differentiability. This paradigm emerged mostly on open-loop datasets, with evaluation focusing not only on driving performance, but also intermediate perception tasks. Unfortunately, architectural advances that excel in open-loop often fail to translate to scalable learning of robust closed-loop driving. In this paper, we systematically re-examine the impact of common architectural patterns on closed-loop performance: (1) high-resolution perceptual representations, (2) disentangled trajectory representations, and (3) generative planning. Crucially, our analysis evaluates the combined impact of these patterns, revealing both unexpected limitations as well as underexplored synergies. Building on these insights, we introduce BevAD, a novel lightweight and highly scalable end-to-end driving architecture. BevAD achieves 72.7% success rate on the Bench2Drive benchmark and demonstrates strong data-scaling behavior using pure imitation learning. Our code and models are publicly available here: https://dmholtz.github.io/bevad/