Augmenting Lane Perception and Topology Understanding with Standard Definition Navigation Maps

TL;DR

This paper introduces a Transformer-based framework that integrates Standard Definition maps to improve real-time lane perception and topology understanding in autonomous driving, offering a scalable alternative to costly HD maps.

Contribution

We propose a novel SD Map Encoder using Transformers to enhance lane topology prediction, significantly boosting accuracy without complex modifications.

Findings

Up to 60% improvement in lane detection accuracy

Effective integration of SD maps into existing methods

Immediate applicability to Transformer-based models

Abstract

Autonomous driving has traditionally relied heavily on costly and labor-intensive High Definition (HD) maps, hindering scalability. In contrast, Standard Definition (SD) maps are more affordable and have worldwide coverage, offering a scalable alternative. In this work, we systematically explore the effect of SD maps for real-time lane-topology understanding. We propose a novel framework to integrate SD maps into online map prediction and propose a Transformer-based encoder, SD Map Encoder Representations from transFormers, to leverage priors in SD maps for the lane-topology prediction task. This enhancement consistently and significantly boosts (by up to 60%) lane detection and topology prediction on current state-of-the-art online map prediction methods without bells and whistles and can be immediately incorporated into any Transformer-based lane-topology method. Code is available at https://github.com/NVlabs/SMERF.