Safety Analysis and Prediction of UAVs Aerial Refueling Docking Based on Deep Learning Data-Driven Method

Abstract

Autonomous aerial refueling (AAR) is essential for both military and civilian applications, but the docking phase poses significant safety risks due to complex environmental conditions that cannot be fully captured by precise mathematical models. This article proposes a data-driven docking predictive model that integrates variational mode decomposition (VMD), sparrow search algorithm (SSA), and long short-term memory (LSTM) neural networks. First, a comprehensive simulation platform for the entire AAR docking system is established to generate reliable data. Then, to address the complex nature of AAR docking signals, VMD decomposes the data into modes with distinct natural frequencies, enhancing input accuracy. SSA optimizes the LSTM parameters, improving prediction accuracy and avoiding local minima. Based on these predictions, a docking safety evaluation network is developed to assess docking safety and prevent failures or collisions. Finally, the performance comparison with other models demonstrates the effectiveness of the proposed approach in diverse scenarios.