Abstract:Acquisition and maintenance are central problems in deploying high-definition (HD) maps for autonomous driving, with two lines of research prevalent in current literature: Online HD map generation and HD map change detection. However, the generated map's quality is currently insufficient for safe deployment, and many change detection approaches fail to precisely localize and extract the changed map elements, hence lacking explainability and hindering a potential fleet-based cooperative HD map update. In this paper, we propose the novel task of explainable element-based HD map change detection and update. In extending recent approaches that use online mapping techniques informed with an outdated map prior for HD map updating, we present ExelMap, an explainable element-based map updating strategy that specifically identifies changed map elements. In this context, we discuss how currently used metrics fail to capture change detection performance, while allowing for unfair comparison between prior-less and prior-informed map generation methods. Finally, we present an experimental study on real-world changes related to pedestrian crossings of the Argoverse 2 Map Change Dataset. To the best of our knowledge, this is the first comprehensive problem investigation of real-world end-to-end element-based HD map change detection and update, and ExelMap the first proposed solution.

What problem does this paper attempt to address?

This paper attempts to solve the problems of updating and maintaining high - definition maps (HD Map) in autonomous driving. Specifically, the paper focuses on: 1. **Deficiencies in existing HD map update methods**: - The quality of the generated maps is not sufficient for safe deployment. - Many change detection methods are unable to accurately locate and extract changed map elements and lack interpretability. 2. **Problems with current evaluation metrics**: - Existing evaluation metrics fail to accurately capture change detection performance and allow for unfair comparisons (for example, comparing map generation methods with and without prior information). 3. **Limitations of datasets**: - Public datasets such as the Argoverse 2 Map Change Dataset only provide a limited number of real - world change scenarios, and most research relies on synthetic data, resulting in a simulation - to - real - world performance gap (sim2real gap). To solve the above problems, the paper proposes a new task - Explainable Element - based HD - Map Change Detection and Update, and introduces a method named ExelMap. The main contributions of this method include: - **Proposing a new task**: Explainable element - based change detection and update to address current challenges. - **Proposing the ExelMap architecture**: This is the first end - to - end explainable element - based HD - map change detection and update architecture. - **Evaluating the deficiencies of existing evaluation strategies**: Discussing the deficiencies of existing evaluation strategies and proposing a comprehensive evaluation metric. ### Specific working principle of ExelMap ExelMap is extended based on LaneSegNet and is implemented through the following steps: 1. **Out - of - date HD map encoder**: Encode the out - of - date HD map into geometric and semantic features. 2. **Encoded prior map integration**: Integrate the encoded prior map features with sensor data through a cross - attention mechanism. 3. **Change detection head**: Add two independent heads for detecting insertion and deletion changes and output the change state of each element. 4. **Dataset and synthetic change generation**: Use the public dataset TbV for training and evaluation, and generate more training samples through synthetic change generation. ### Evaluation metrics To evaluate the effect of ExelMap, the paper designs nine evaluation metrics, covering different types of change detection, localization accuracy, and the AP score of the updated map. These metrics ensure a comprehensive evaluation of the model performance, especially for the detection and localization ability of change types. ### Conclusion Through ExelMap, the paper not only solves the deficiencies of existing HD map update methods but also provides new directions and evaluation criteria for future research. This helps to improve the accuracy and reliability of HD maps in autonomous driving systems, thereby ensuring driving safety.

ExelMap: Explainable Element-based HD-Map Change Detection and Update

Mind the map! Accounting for existing map information when estimating online HDMaps from sensor

Exploring Real World Map Change Generalization of Prior-Informed HD Map Prediction Models

High Definition Map Mapping and Update: A General Overview and Future Directions

Real-Time HD Map Change Detection for Crowdsourcing Update Based on Mid-to-High-End Sensors

Simultaneous Localization and Map Change Update for the High Definition Map-Based Autonomous Driving Car

Vision-HD: road change detection and registration using images and high-definition maps

HoMap: End-to-End Vectorized HD Map Construction with High-order Modeling

HDMapNet: An Online HD Map Construction and Evaluation Framework

Tutorial on High-Definition Map Generation for Automated Driving in Urban Environments

EdgeMap: CrowdSourcing High Definition Map in Automotive Edge Computing

ScalableMap: Scalable Map Learning for Online Long-Range Vectorized HD Map Construction

VectorMapNet: End-to-end Vectorized HD Map Learning

EAN-MapNet: Efficient Vectorized HD Map Construction with Anchor Neighborhoods

HDMapNet: A Local Semantic Map Learning and Evaluation Framework.

Enhancing Online Road Network Perception and Reasoning with Standard Definition Maps

Enhancing Vectorized Map Perception with Historical Rasterized Maps

HIMap: HybrId Representation Learning for End-to-end Vectorized HD Map Construction

A Review on End-to-End High-Definition Map Generation

HD Map Generation from Noisy Multi-Route Vehicle Fleet Data on Highways with Expectation Maximization

A Review of Crowdsourcing Update Methods for High-Definition Maps