Low-speed UAV trajectory prediction under short-term data loss: a preliminary feasibility study for vehicle-mounted...

Area of Science:

• Robotics and Control Systems
• Computer Vision
• Aerospace Engineering

Background:

• Short-term data loss in electro-optical (EO) tracking systems disrupts the continuity and stability of monitoring unmanned aerial vehicles (UAVs).
• Accurate UAV trajectory prediction is crucial for maintaining effective tracking during these data-loss periods.

Purpose of the Study:

• To develop and evaluate a novel UAV trajectory prediction method specifically designed for short-term data-loss scenarios in EO tracking.
• To improve the robustness and accuracy of UAV tracking systems when encountering intermittent signal interruptions.

Main Methods:

• Utilized an improved Support Vector Machine (SVM) for accurate identification of UAV flight states based on trajectory data.
• Developed a joint architecture combining Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks for temporal feature extraction.
• Integrated an adaptive attention mechanism to enhance the learning of critical trajectory features, guided by recognized flight states.

Main Results:

• The improved SVM achieved over 90% accuracy in recognizing basic UAV flight states.
• The proposed method demonstrated superior performance in root mean square error (RMSE) for turning flight prediction compared to LSTM and IMM-EKF under 2s data loss.
• Achieved comparable accuracy to TCN and IMM-EKF for non-turning states, with a lower computational cost and fewer parameters than lightweight TCN.

Conclusions:

• The proposed algorithm effectively addresses short-term data loss challenges in vehicle-mounted EO tracking of low-speed, mildly maneuverable UAVs.
• Flight-state recognition and state-guided attention mechanisms significantly enhance trajectory prediction precision.
• The method shows favorable overall performance across various conditions, defining its practical application scope.