LiDAR- and Radar-Based Robust Vehicle Localization with Confidence Estimation of Matching Results.

Area of Science:

• Robotics
• Computer Vision
• Sensor Fusion

Background:

• Accurate localization is critical for safe autonomous driving.
• Current map-matching methods using road surface patterns fail in environments with obscured features, such as snow cover.
• Millimeter-wave radar offers potential for localization in adverse weather but lacks the precision of LiDAR.

Purpose of the Study:

• To develop a robust localization system for autonomous vehicles that functions reliably in both clear and snowy conditions.
• To overcome the limitations of environmental factors affecting traditional map-matching localization techniques.

Main Methods:

• A sensor fusion approach combining Light Detection and Ranging (LiDAR) and millimeter-wave radar.
• A system prioritizing LiDAR-based localization when road surface patterns are visible.
• A system emphasizing millimeter-wave radar-based localization when road surface patterns are obscured by snow or other environmental factors.

Main Results:

• The proposed hybrid system achieves accurate and robust position estimation across diverse environmental conditions.
• Successful autonomous driving was maintained irrespective of whether the road surface was visible or obscured by snow.
• The system demonstrated improved resilience compared to methods relying solely on visual or LiDAR-based features.

Conclusions:

• Combining LiDAR and millimeter-wave radar significantly enhances localization robustness for autonomous driving in challenging environments.
• The adaptive fusion strategy ensures continuous and reliable positioning, crucial for safety-critical applications.
• This approach addresses a key limitation in current autonomous driving localization technologies, particularly in adverse weather.