HD Map Construction

HD: High-Definition · Also known as MAP

The field builds detailed road maps—such as lane boundaries, dividers, and connectivity—from vehicle cameras, LiDAR, radar, aerial imagery, and geographic data, supporting navigation, planning, and simulation. Common approaches use bird’s-eye-view transformers to predict vectorized map elements, with sensor fusion, geometric priors, and temporal consistency to improve structure across views and drives. Research also explores crowdsourced map aggregation, semi-supervised learning, and generative models, with nuScenes, Argoverse 2, and OpenLane-V2 among the evaluation settings. A central challenge is maintaining accurate geometry and topology despite localization errors, missing or changing sensor views, and geographic variation, while scaling reliable maps beyond individual vehicle observations.

Latest papers

39 papers5 papers in the last four weeks, up 67% on the four weeks before. 0.0% of all new papers.

Failure Modes for Deep Learning-Based Online Mapping: How to Measure and Address Them
Mar 20, 2026Michael Hubbertz, Qi Han, Tobias MeisenOOD Generalization HD Map Construction
University of Wuppertal · Aptiv Services Deutschland GmbH
MapGCLR: Geospatial Contrastive Learning of Representations for Online Vectorized HD Map Construction
Mar 11, 2026Jonas Merkert, Alexander Blumberg, Jan-Hendrik Pauls +1HD Map Construction Geospatial Representation Learning
Institute of Measurement and Control Systems, Karlsruhe Institute of Technology (KIT), Karlsruhe, Germany
Radar-based Pose Optimization for HD Map Generation from Noisy Multi-Drive Vehicle Fleet Data
Mar 3, 2026Alexander Blumberg, Jonas Merkert, Christoph StillerHD Map Construction Simultaneous Localization and Mapping
Institute of Measurement and Control Systems, Karlsruhe Institute of Technology (KIT), Karlsruhe, Germany
Impact of Localization Errors on Label Quality for Online HD Map Construction
Mar 3, 2026Alexander Blumberg, Jonas Merkert, Richard Fehler +4HD Map Construction Autonomous Driving Datasets
Institute of Measurement and Control Systems, Karlsruhe Institute of Technology (KIT), Karlsruhe, Germany · FZI Research Center for Information Technology, Karlsruhe, Germany
B $^2$ F-Map: Crowd-sourced Mapping with Bayesian B-spline Fusion
Mar 2, 2026Yiping Xie, Yuxuan Xia, Erik Stenborg +5HD Map Construction Bayesian Filtering
Zenseact, Lindholmspiren 2, Gothenburg, Sweden. · Department of Electrical Engineering, Linköping University, Linköping, Sweden. · Yuxuan Xia was a Post-doc with Zenseact, and he is now with the Department of Automation and Perception, Shanghai Jiao Tong University, Shanghai, China. +1
SEF-MAP: Subspace-Decomposed Expert Fusion for Robust Multimodal HD Map Prediction
Feb 25, 2026Haoxiang Fu, Lingfeng Zhang, Hao Li +7Camera-LiDAR Fusion Multimodal Robustness
National University of Singapore · Shenzhen Tsinghua International Graduate School, Tsinghua Univeristy · Xiaomi EV +4
PredMapNet: Future and Historical Reasoning for Consistent Online HD Vectorized Map Construction
Feb 18, 2026Bo Lang, Nirav Savaliya, Zhihao Zheng +3HD Map Construction
Lehigh University · Honda Research Institute USA
FlexMap: Robust HD Map Construction under Flexible Camera Configurations
Jan 29, 2026Run Wang, Chaoyi Zhou, Amir Salarpour +5HD Map Construction Autonomous Driving
School of Computing, Clemson University · 2Tacoma School of Engineering and Technology, University of Washington
L2G-Map: Local-to-Global Mapping via Hierarchical Diffusion Refinement and Elliptical Bayesian Fusion
Mar 4, 2025Siyu Li, Xinying Hong, Fei Teng +5HD Map Construction Autonomous Driving
School of Automation and Electrical Engineering, Zhejiang University of Science and Technology, China · School of Artificial Intelligence and Robotics and the National Engineering Research Center of Robot Visual Perception and Control Technology, Hunan University, China · College of Computer Science and Electronic Engineering, Hunan University, China +1

Mar 20, 2026J/K move · Enter open · S save

Failure Modes for Deep Learning-Based Online Mapping: How to Measure and Address Them

Michael Hubbertz, Qi Han, Tobias Meisen

University of Wuppertal · Aptiv Services Deutschland GmbH

OOD Generalization HD Map Construction Autonomous Driving Datasets Autonomous Driving Data Selection

Open paper AskPDF arXiv

Deep learning-based online mapping has emerged as a cornerstone of autonomous driving, yet these models frequently fail to generalize beyond familiar environments. We propose a framework to identify and measure the underlying failure modes by disentangling two effects: Memorization of input features and overfitting to known map geometries. We propose measures based on evaluation subsets that control for geographical proximity and geometric similarity between training and validation scenes. We introduce Fréchet distance-based reconstruction statistics that capture per-element shape fidelity without threshold tuning, and define complementary failure-mode scores: a localization overfitting score quantifying the performance drop when geographic cues disappear, and a map geometry overfitting score measuring degradation as scenes become geometrically novel. Beyond models, we analyze dataset biases and contribute map geometry-aware diagnostics: A minimum-spanning-tree (MST) diversity measure for training sets and a symmetric coverage measure to quantify geometric similarity between splits. Leveraging these, we formulate an MST-based sparsification strategy that reduces redundancy and improves balancing and performance while shrinking training size. Experiments on nuScenes and Argoverse 2 across multiple state-of-the-art models yield more trustworthy assessment of generalization and show that map geometry-diverse and balanced training sets lead to improved performance. Our results motivate failure-mode-aware protocols and map geometry-centric dataset design for deployable online mapping.