Artificial Marker and MEMS IMU-Based Pose Estimation Method to Meet Multirotor UAV Landing Requirements

Area of Science:

• Robotics and Automation
• Computer Vision
• Navigation Systems

Background:

• Autonomous operation of multirotor unmanned air vehicles (UAVs) necessitates precision landing capabilities.
• Fiducial markers and onboard cameras are common, cost-effective solutions for this critical phase.
• Existing systems can be unreliable during marker occlusion or poor lighting conditions.

Purpose of the Study:

• To propose a six-degrees-of-freedom (DoF) pose estimation solution for UAV precision landing.
• To integrate artificial marker detection with micro-electromechanical system (MEMS) inertial measurement unit (IMU) data.
• To enhance positioning reliability and accuracy, especially during temporary vision loss.

Main Methods:

• A monocular camera detects an artificial marker and extracts corner points for absolute UAV positioning and heading estimation.
• Data from a MEMS IMU is fused with visual data using an extended Kalman filter (EKF) for continuous positioning.
• IMU sensor error terms are modeled and estimated to improve fusion accuracy.

Main Results:

• The proposed system achieves centimeter-level positioning accuracy and sub-0.1-degree heading error.
• Roll and pitch angle errors were reduced by an average of 33% and 54%, respectively, compared to marker-only approaches.
• During a five-second vision outage, horizontal and vertical position drifts were 0.41 m and 0.09 m, respectively.

Conclusions:

• The integrated marker-IMU system provides robust and accurate pose estimation for UAV precision landing.
• The EKF-based sensor fusion effectively mitigates issues caused by temporary marker occlusion.
• The system demonstrates significant improvements in landing accuracy and stability, crucial for autonomous drone applications.