Landing Point Prediction of Aircraft Based on Transfer Learning Method

The prediction of aircraft landing points is directly related to the safety of the aircraft, the success of the mission, and the efficiency of search and rescue operations. Traditional methods for predicting aircraft landing points include numerical prediction methods and analytical prediction methods. The former uses onboard computers to integrate the aircraft's motion equations in real-time to predict the landing point, offering strong adaptability but involving substantial computational load. The latter searches for approximate analytical solutions from all or part of the flight trajectories, requiring less ballistic computation but resulting in lower prediction accuracy. Machine learning can train and optimize models using large amounts of historical flight data and real-time observation data. This paper first briefly outlines the current research status of aircraft landing point prediction, and then studies various machine learning and deep learning models for predicting aircraft landing points. Finally, to overcome the small sample problem, the paper attempts to introduce transfer learning to enhance model performance and application range, and the effectiveness and superiority of machine learning and transfer learning models are verified and compared using large-scale simulation platform data.