Integration of a Laser Precision Landing System for UAVs in All-Weather and Dynamic Environments

Abstract

Autonomous unmanned aerial vehicle (UAV) landings in GPS-denied and visually degraded environments remain a significant challenge. Standard Global Navigation Satellite System (GNSS) solutions are susceptible to multipath interference near metallic structures and provide absolute rather than relative positioning, making them unsuitable for landing on dynamic platforms. Passive visual sensors fail under low light, fog, and direct glare, and introduce significant processing latency. This thesis integrates an active infrared precision landing system based on the Sixdof Space Kipa optical sensor, which operates at up to 400Hz across all lighting conditions. A decentralized control architecture was developed combining a LattePanda Mu companion computer with a Pixhawk 6C flight controller. A custom C++ application implements a three-stage signal processing pipeline (median filter, chi-squared outlier gate, 1D Kalman filter), weather-adaptive proportional-integral-derivative control with gain scheduling, a hierarchical descent governor, and a finite state machine enforcing safe behavior under sensor loss and extreme conditions. Velocity commands are transmitted to the flight controller via the MAVLink protocol. Simulation-in-the-loop across 24 test configurations demonstrated sub-centimeter mean landing error under calm conditions, Excellent classification (landing accuracy within 300mm) at steady wind speeds up to 15m/s, and Excellent classification at turbulence intensities up to 7. Under a combined worst-case scenario of 15m/s wind and turbulence intensity 5, all five runs confirmed touchdown with a mean error of 181.4mm. The end-to-end pipeline latency was measured at 52ms, satisfying the 75ms design objective. Physical characterization of the Kipa sensor confirmed sub 3cm depth accuracy at touchdown range and near-zero position noise. Hardware bench integration was verified through confirmed motor spin-up under algorithmic control.