Vision-Aided Velocity Estimation in GNSS Degraded or Denied Environments

Area of Science:

• Robotics
• Navigation Systems
• Computer Vision

Background:

• Global Positioning System (GPS) availability is often unreliable in remote piloting scenarios, impacting vehicle navigation and control.
• Accurate velocity estimation is critical for maintaining vehicle guidance and control performance, especially during GPS outages.

Purpose of the Study:

• To develop a novel navigation system architecture for robust velocity estimation in environments with intermittent GPS.
• To ensure reliable vehicle guidance and control by maintaining accurate velocity estimates.

Main Methods:

• A novel landmark-extraction algorithm identifies environmental features for precise velocity adjustments.
• A Sequential Kalman filter processes camera data relative to landmarks for state estimation.
• LiDAR odometry is integrated with landmark tracking to minimize velocity error.

Main Results:

• The system demonstrated effective velocity estimation capabilities in real-world trajectory tests.
• Performance was validated under diverse conditions, including scenarios with compromised or absent GPS signals.
• The proposed architecture successfully maintained stable velocity estimation.

Conclusions:

• The developed navigation system provides a feasible solution for remote piloting in GPS-denied environments.
• The system's ability to accurately estimate velocity enhances vehicle control reliability.
• This research contributes to robust autonomous navigation in challenging operational conditions.