Design of a 3D High-Definition Map Visualizer for Pose Estimation and Autonomous Navigation in Dynamic Environments.

Area of Science:

• Robotics
• Computer Vision
• Autonomous Systems

Background:

• Autonomous navigation relies on accurate high-definition (HD) maps and reliable sensor data.
• Existing systems face challenges in dynamic environments and with sensor degradation, particularly GNSS-denied areas.

Purpose of the Study:

• To develop and validate an HD map development framework with real-time visualization for autonomous navigation.
• To enhance state estimation and motion planning using multimodal perception data.
• To ensure data reliability through sensor health monitoring.

Main Methods:

• Integration of synchronized visual and LiDAR data for HD map generation.
• Development of a semantic-aware visual odometry (VO) module for pose estimation.
• Implementation of a sensor health monitoring system for data reliability.

Main Results:

• The framework successfully generates accurate and interpretable HD maps for dynamic environments.
• The semantic-aware VO module achieves reliable pose estimation, even under degraded perceptual conditions.
• The sensor health monitoring system validates data integrity in challenging urban scenarios.

Conclusions:

• The proposed HD map visualizer and perception modules are transferable for autonomous navigation.
• The framework serves as an effective benchmarking tool for state estimation and motion planning algorithms.
• The system demonstrates robustness and reliability for autonomous driving applications.