UWB based drone localization for autonomous landing systems

Accurate localization is essential for autonomous unmanned aerial vehicle (UAV) operations, particularly during critical maneuvers such as landing. This thesis presents a robust, high-accuracy localization system utilizing Ultra-Wideband (UWB) technology combined with sensor fusion. The proposed system employs UWB anchors arranged compactly on a 0.6 m × 0.7 m platform and a customized UWB tag equipped with inertial orientation sensors for precise positioning. Real-time data processing, including sensor calibration, signal filtering, and fusion, is implemented on a Raspberry Pi 5 microcomputer. Orientation estimation is performed using the Versatile Quaternion-based Filter (VQF) algorithm, which integrates calibrated data from accelerometer, gyroscope, and magnetometer sensors. To enhance distance measurement accuracy, a multilayer perceptron (MLP) model is trained and applied, significantly reducing the UWB ranging error from 9.26 cm to 4.35 cm root mean square error (RMSE). An Extended Kalman Filter (EKF) further integrates these corrected UWB measurements with dynamic acceleration and orientation data to achieve resilient and accurate localization. Experimental evaluations confirm the system's high performance, demonstrating an average RMSE of 4.28 cm in static scenarios. Dynamic tests validate reliable tracking capabilities across diverse trajectories, underscoring the system's suitability for autonomous UAV landing applications. The presented research provides a solid groundwork for future developments in sensor calibration, data integration, and advanced localization algorithms.