Free-Resolution Probability Distributions Map-Based Precise Vehicle Localization in Urban Areas.

Area of Science:

• Robotics
• Computer Vision
• Geospatial Analysis

Background:

• Accurate vehicle localization is crucial for autonomous driving and advanced driver-assistance systems.
• Existing probability distribution maps often face limitations in resolution flexibility and data storage efficiency.
• 3D Light Detection and Ranging (LIDAR) provides rich environmental data for localization.

Purpose of the Study:

• To propose a novel Free-Resolution Probability Distributions Map (FRPDM) for enhanced vehicle localization.
• To develop an FRPDM-based precise localization method using 3D LIDAR data.
• To evaluate the efficiency and accuracy of the proposed method in a complex urban setting.

Main Methods:

• FRPDM generation using Gaussian mixture modeling on road markings and vertical structures from 3D LIDAR point clouds.
• FRPDM-based localization employing point-to-probability distribution scan matching and feature-point matching.
• Vehicle position and heading estimation using an extended Kalman filter-based navigation filter.

Main Results:

• The proposed FRPDM achieved a very small map size (61 KB/km) due to efficient object representation.
• Experimental results in Seoul's Gangnam area demonstrated high localization accuracy.
• Root Mean Square (RMS) position errors were 0.057 m (lateral) and 0.178 m (longitudinal).

Conclusions:

• The FRPDM offers an efficient and effective representation for 3D LIDAR-based mapping.
• The FRPDM-based localization method achieves precise vehicle positioning and heading estimation.
• The approach shows significant promise for real-world autonomous navigation applications.