Neural Data Fusion Enhanced PD Control for Precision Drone Landing in Synthetic Environments

Unmanned aerial vehicles are increasingly used in applications like surveillance, mapping, and delivery, where precise and safe landings are crucial, especially on small, designated platforms in multimodal systems. However, achieving consistent landing accuracy is challenging due to sensor inaccuracies, environmental disturbances, and dynamic conditions. This paper introduces a synthetic drone simulation environment to test landing scenarios and control strategies. We propose a control framework combining a classical PD controller with a neural network-based data fusion model that corrects sensor noise, particularly in GPS and UWB data, to improve landing precision. Simulation results demonstrate that our approach enhances landing accuracy and robustness compared to traditional methods.