Flight Trajectory Prediction Based on Automatic Dependent Surveillance-Broadcast Data Fusion with Interacting...

Area of Science:

• Aerospace Engineering
• Artificial Intelligence
• Data Science

Background:

• Aircraft trajectory prediction faces challenges due to uncertain motion and complex temporal dependencies in flight data.
• Existing methods struggle with the dynamic and unpredictable nature of flight paths.

Purpose of the Study:

• To develop a novel hybrid framework, the IMM-Informer, for enhanced aircraft trajectory prediction.
• To improve the accuracy and robustness of flight path forecasting by integrating multiple models.

Main Methods:

• Proposed a hybrid framework integrating Interacting Multiple Model (IMM) with a deep learning Informer model.
• Utilized IMM for initial state predictions and Informer for trajectory deviation correction.
• Employed ProbSparse self-attention mechanism within the Informer encoder for temporal feature extraction.

Main Results:

• The IMM-Informer demonstrated notable reductions in prediction errors.
• Achieved significantly higher prediction accuracies compared to standalone sequence prediction models.
• Validated effectiveness using real-world Automatic Dependent Surveillance-Broadcast (ADS-B) sensor data.

Conclusions:

• The IMM-Informer framework offers a robust and accurate solution for aircraft trajectory prediction.
• Hybrid approaches combining traditional methods with deep learning show significant promise.
• The method effectively addresses the complexities of flight dynamics and temporal data dependencies.