Crowd-Sourced Mapping of New Feature Layer for High-Definition Map

Area of Science:

• Intelligent Transportation Systems
• Geomatics Engineering
• Computer Vision

Background:

• High-Definition maps (HD maps) are crucial for intelligent vehicles, requiring continuous updates with new feature layers for enhanced performance in complex environments.
• Traditional HD map generation is expensive and time-consuming due to the need for professional mapping vehicles to re-survey all public roads.

Purpose of the Study:

• To propose a cost-effective, crowd-sourced mapping process for generating new feature layers for HD maps.
• To reduce the financial and temporal costs associated with updating HD maps.

Main Methods:

• A two-step crowd-sourced mapping process is introduced, utilizing data from multiple intelligent vehicles.
• New environmental features are acquired by intelligent vehicles and used to build individual feature layers via the HD map-based GraphSLAM approach.
• These crowd-sourced feature layers are transmitted to a central map cloud via mobile networks for integration.

Main Results:

• Simulations were conducted to analyze and evaluate the performance of the proposed crowd-sourced mapping process.
• Real-world driving experiments validated the simulation findings, confirming the effectiveness of the approach.

Conclusions:

• The proposed crowd-sourced mapping process offers a viable and efficient solution for the continuous updating of HD maps.
• This method significantly reduces the costs associated with generating new feature layers, facilitating improved intelligent vehicle performance.