Attention-LSTM based prediction model for aircraft 4-D trajectory.

Area of Science:

• Aviation safety
• Artificial intelligence in aviation
• Air traffic management

Background:

• Increasing global aviation activities necessitate enhanced airspace capacity and safety.
• Accurate aircraft trajectory prediction is crucial for air traffic safety but challenged by sparse way-points and flight uncertainties.
• Existing trajectory prediction models face difficulties in handling complex flight dynamics.

Purpose of the Study:

• To propose an advanced trajectory prediction model to improve accuracy in civil aviation.
• To address the challenge of predicting aircraft trajectories amidst limited airspace and uncertain flight conditions.
• To enhance the reliability of air traffic safety through superior trajectory prediction.

Main Methods:

• Developed an attention-Long Short-Term Memory (LSTM) neural network model for trajectory prediction.
• Utilized LSTM for extracting time-series features from flight trajectory data.
• Applied an attention mechanism to prioritize significant features and reduce the impact of secondary elements.

Main Results:

• The proposed attention-LSTM model demonstrated superior performance compared to traditional models like LSTM, SVM, BP, HMM, and CNN-LSTM.
• Quantitative analysis confirmed the enhanced accuracy of the attention-LSTM model in trajectory prediction.
• The attention mechanism effectively improved the model's ability to focus on critical trajectory information.

Conclusions:

• The attention-LSTM model offers a significant advancement in aircraft trajectory prediction accuracy.
• This improved prediction capability can contribute to safer and more efficient air traffic management.
• The model's ability to handle flight uncertainties makes it a valuable tool for future aviation systems.