Adaptive Data Fusion Method of Multisensors Based on LSTM-GWFA Hybrid Model for Tracking Dynamic Targets

Area of Science:

• Robotics and Autonomous Systems
• Aerospace Engineering
• Artificial Intelligence

Background:

• The increasing complexity of future battlefields necessitates advanced target tracking capabilities.
• Unmanned Aerial Vehicles (UAVs) equipped with multisensors are crucial for real-time dynamic target monitoring.
• Traditional multisensor fusion algorithms face challenges with temporal data correlations and target maneuvering.

Purpose of the Study:

• To develop an improved data fusion method for dynamic target tracking using UAVs.
• To enhance the accuracy and robustness of target tracking in air combat scenarios.
• To validate the proposed method through simulations and field experiments.

Main Methods:

• A novel data fusion method combining a global Kalman filter with Long Short-Term Memory (LSTM) prediction for measurement variance.
• Implementation of an adaptive truncation mechanism for optimal weight determination.
• Inclusion of a detection mechanism for maneuvering targets and consideration of temporal data correlations.

Main Results:

• Numerical simulations demonstrated a 66% improvement in tracking accuracy using the proposed algorithm with 871 flight data.
• Field experiments on a refitted civil wing UAV platform confirmed the method's effectiveness and stability.
• The data fusion method exhibited strong generalization ability in real-world dynamic target tracking scenarios.

Conclusions:

• The developed data fusion method significantly enhances the accuracy of dynamic target tracking for UAVs.
• The integration of LSTM prediction and adaptive weighting provides a robust solution for complex tracking scenarios.
• The method's validated effectiveness, stability, and generalization ability make it suitable for future military applications.