Novel Intersection Type Recognition for Autonomous Vehicles Using a Multi-Layer Laser Scanner.

Area of Science:

• Robotics
• Computer Vision
• Autonomous Systems

Background:

• Autonomous vehicles (AVs) require accurate intersection recognition for safe navigation.
• Urban environments present diverse intersection types (merge-roads, diverge-roads, plus-shape, T-shape).
• Existing methods may lack robustness in complex, real-world scenarios.

Purpose of the Study:

• To propose a novel method for intersection type recognition for autonomous vehicles.
• To enhance AV safety by enabling accurate identification of diverse intersection geometries.
• To utilize multi-layer laser scanner data for robust intersection classification.

Main Methods:

• A two-step approach: (1) Static Local Coordinate Occupancy Grid Map (SLOGM) building using a dynamic binary Bayes filter.
• (2) Intersection classification using the generated SLOGM as input features.
• Leveraging multi-layer laser scanner data for map generation and classification.

Main Results:

• The proposed method successfully builds SLOGMs relative to the local coordinate system.
• Experimental validation in a real-world environment demonstrates the method's effectiveness.
• The SLOGM serves as a reliable attribute for accurate intersection classification.

Conclusions:

• The novel SLOGM-based method provides a robust solution for intersection type recognition in autonomous driving.
• This approach contributes to improved safety and reliability of autonomous vehicles in urban settings.
• The method's validity is confirmed through practical experimentation.