Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GlobalMapNet: An Online Framework for Vectorized Global HD Map Construction

Anqi Shi, Yuze Cai, Xiangyu Chen, Jian Pu, Zeyu Fu, Hong Lu

TL;DR

This paper introduces GlobalMapNet, the first online framework for vectorized global HD map construction, which updates and utilizes a global map on the ego vehicle, and designs a new algorithm, Map NMS, to remove duplicate map elements and produce a clean map.

Abstract

High-definition (HD) maps are essential for autonomous driving systems. Traditionally, an expensive and labor-intensive pipeline is implemented to construct HD maps, which is limited in scalability. In recent years, crowdsourcing and online mapping have emerged as two alternative methods, but they have limitations respectively. In this paper, we provide a novel methodology, namely global map construction, to perform direct generation of vectorized global maps, combining the benefits of crowdsourcing and online mapping. We introduce GlobalMapNet, the first online framework for vectorized global HD map construction, which updates and utilizes a global map on the ego vehicle. To generate the global map from scratch, we propose GlobalMapBuilder to match and merge local maps continuously. We design a new algorithm, Map NMS, to remove duplicate map elements and produce a clean map. We also propose GlobalMapFusion to aggregate historical map information, improving consistency of prediction. We examine GlobalMapNet on two widely recognized datasets, Argoverse2 and nuScenes, showing that our framework is capable of generating globally consistent results.

GlobalMapNet: An Online Framework for Vectorized Global HD Map Construction

TL;DR

This paper introduces GlobalMapNet, the first online framework for vectorized global HD map construction, which updates and utilizes a global map on the ego vehicle, and designs a new algorithm, Map NMS, to remove duplicate map elements and produce a clean map.

Abstract

High-definition (HD) maps are essential for autonomous driving systems. Traditionally, an expensive and labor-intensive pipeline is implemented to construct HD maps, which is limited in scalability. In recent years, crowdsourcing and online mapping have emerged as two alternative methods, but they have limitations respectively. In this paper, we provide a novel methodology, namely global map construction, to perform direct generation of vectorized global maps, combining the benefits of crowdsourcing and online mapping. We introduce GlobalMapNet, the first online framework for vectorized global HD map construction, which updates and utilizes a global map on the ego vehicle. To generate the global map from scratch, we propose GlobalMapBuilder to match and merge local maps continuously. We design a new algorithm, Map NMS, to remove duplicate map elements and produce a clean map. We also propose GlobalMapFusion to aggregate historical map information, improving consistency of prediction. We examine GlobalMapNet on two widely recognized datasets, Argoverse2 and nuScenes, showing that our framework is capable of generating globally consistent results.
Paper Structure (15 sections, 6 equations, 3 figures, 6 tables)

This paper contains 15 sections, 6 equations, 3 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Relate works
  3. Global Map Construction
  4. Task Formulation
  5. Evaluating global map construction
  6. GlobalMapNet
  7. Local mapping
  8. GlobalMapBuilder
  9. GlobalMapFusion
  10. Experiments
  11. Implementation details
  12. Results
  13. Ablation studies
  14. Visualization
  15. Conclusion

Figures (3)

  • Figure 1: The relationship and difference between local map construction and global map construction. In global map construction, multi-run local mapping results are merged sequentially to produce the global map.
  • Figure 2: The structure of GlobalMapNet. Our method consists of an online local mapping system, the GlobalMapBuilder and the GlobalMapFusion. The global map is kept in permanent storage and updated continuously with local map predictions. Historical map prior is fused to produce consistent local maps, forming closed-loop global map construction and utilization.
  • Figure 3: Visualized single-scene results on two datasets. a) nuScenes: GlobalMapNet performs better in predicting the road intersection (yellow circles). Both methods fail to generate tangled road boundaries (purple circles). b) Argoverse2: GlobalMapNet generates a continuous road boundary, while StreamMapNet fails with a broken prediction (yellow circles). Both methods fail to predict complicated lane dividers (purple circles).