Vision-Based SLAM System for Unmanned Aerial Vehicles

Area of Science:

• Robotics
• Computer Vision
• Navigation Systems

Background:

• Unmanned Aerial Vehicles (UAVs) require accurate localization and mapping for autonomous operation.
• Existing systems often rely heavily on GPS, which can be unreliable or unavailable in certain environments.
• Sensor fusion is crucial for improving the robustness and accuracy of UAV navigation.

Purpose of the Study:

• To develop a novel vision-based simultaneous localization and mapping (SLAM) system for UAVs.
• To propose an Extended Kalman Filter-based estimator for fusing data from multiple sensors.
• To enable robust, vision-based navigation for UAVs, reducing reliance on GPS.

Main Methods:

• Implementation of an Extended Kalman Filter (EKF) for state estimation.
• Fusion of measurements from an Attitude and Heading Reference System (AHRS), GPS, and a monocular camera.
• Utilizing landmarks for metric scale recovery and vision-only navigation post-initialization.

Main Results:

• The proposed estimator successfully fuses data from AHRS, GPS, and monocular camera.
• Experimental results demonstrate significant improvement in trajectory estimation accuracy compared to GPS-only methods.
• The system achieves accurate localization and mapping using vision-based data, even when GPS is unavailable.

Conclusions:

• The novel EKF-based estimator effectively integrates vision data for enhanced UAV localization and mapping.
• Vision-based navigation significantly improves trajectory estimation accuracy and robustness for UAVs.
• The proposed system offers a viable solution for reliable UAV navigation in GPS-denied environments.