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

Area of Science:

• Robotics and Intelligent Systems
• Geographic Information Systems
• Computer Vision

Background:

• High-definition (HD) maps are essential for advanced automated driving (AD) systems, particularly for levels 4 and higher.
• HD maps provide crucial environmental data beyond the limitations of conventional vehicle sensors.
• These maps typically consist of pointcloud data for localization and vector data for path planning.

Purpose of the Study:

• To introduce and detail two distinct, sequential workflows for generating HD maps.
• To highlight the critical roles of registration and mapping processes in creating pointcloud and vector maps, respectively.
• To validate the generated HD maps using established AD software stacks.

Main Methods:

• Utilized open-source software (OSS) tools, specifically CloudCompare and ASSURE, for HD map generation.
• Developed and documented two successive HD map generation workflows, focusing on registration and mapping stages.
• Integrated and validated the generated HD maps within the Autoware OSS stack for AD systems.

Main Results:

• Successfully generated HD maps using specified OSS tools and workflows.
• Validated the efficacy of the generated HD maps through integration with localization and path-planning modules in Autoware.
• Demonstrated the capability of HD maps to support environmental-monitoring vehicles operating at level 4 autonomy.

Conclusions:

• The presented HD map generation workflows are effective and viable for creating detailed maps for automated driving.
• The use of OSS tools like CloudCompare, ASSURE, and Autoware facilitates the development and validation of HD mapping technologies.
• Generated HD maps are proven to be functional for enabling level 4 automated driving capabilities.