Scale Factor Estimation for Quadrotor Monocular-Vision Positioning Algorithms.

Area of Science:

• Robotics
• Control Systems
• Computer Vision

Background:

• Autonomous navigation for unmanned aerial vehicles (UAVs) requires accurate state estimation.
• Existing methods often rely on multiple sensors due to limitations of single-sensor systems.
• Monocular vision-based simultaneous localization and mapping (SLAM) presents challenges in determining accurate position scale factors.

Purpose of the Study:

• To develop a deterministic estimator for reconstructing the position scale factor in monocular SLAM for quadrotor UAVs.
• To design a deterministic observer for estimating quadrotor translational velocity using only onboard sensor data.
• To validate the proposed methods using inertial and visual measurements.

Main Methods:

• Implementation of a deterministic estimator and observer based on the immersion and invariance (I&I) method.
• Fusion of inertial measurement unit (IMU) and visual odometry data.
• Utilization of Lyapunov stability theory to prove convergence of estimation errors.

Main Results:

• The proposed estimator successfully reconstructs the unknown position scale factor for monocular SLAM.
• The observer accurately estimates the quadrotor's translational velocity.
• Numerical simulations confirm the asymptotic convergence of estimation errors to zero.

Conclusions:

• The developed deterministic approach enhances the accuracy of UAV state estimation.
• This method enables more reliable autonomous navigation for quadrotors using monocular vision and IMU data.
• The validated techniques contribute to the advancement of robust UAV navigation systems.