Aircraft Intent Prediction Using Continuous Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Trajectory Predictors for Arrival Management are limited in their ability to accurately predict aircraft trajectories due to intrinsic prediction errors, which are linked to their architecture and limitations in describing aircraft intent, leading to reduced efficiency in resource utilization and precision in future traffic evolution predictions.
Innovation Solution
A method utilizing a continuous learning process based on Aircraft Intent Description Language (AIDL) and historical flight data, which optimizes aircraft intent parameters through an optimization algorithm to minimize prediction errors, thereby improving the accuracy of trajectory predictions for specific aircraft types, airlines, and weather conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a static aircraft intent description is used for all aircraft of the same model under identical conditions, then the device complexity is reduced and ease of operation is improved, but prediction precision deteriorates due to intrinsic prediction errors
Solution Approach 1:
The patent transforms the static aircraft intent description into a dynamic one by introducing a learning process that continuously adapts parameters based on historical flight data. The system evolves from fixed predetermined values to dynamically adjusted parameters that reflect actual aircraft behavior patterns, thereby improving prediction precision without requiring a fundamentally more complex system architecture.
Solution Approach 2:
The patent applies parameter changes by optimizing specific aircraft intent parameters (such as descent rate, speed profiles, altitude adjustments) based on historical data analysis. Instead of changing the overall system structure, the invention fine-tunes individual parameters to reduce prediction errors, achieving higher precision while maintaining the existing trajectory description framework.
2Measurement precision
If historical flight data and continuous learning process are implemented to reduce prediction errors, then prediction precision is improved, but device complexity and computational requirements increase
Solution Approach 1:
The patent implements preliminary action by pre-processing and storing historical flight data in structured formats before actual prediction operations. The system performs offline analysis to extract behavioral patterns and stores optimized parameter sets that can be quickly applied during real-time predictions, reducing the computational burden during operational phases while maintaining high prediction accuracy.
Solution Approach 2:
The system applies self-service by automatically learning from historical data and self-optimizing its prediction parameters without requiring manual intervention. The continuous learning process autonomously identifies patterns, adjusts parameters, and improves prediction capabilities over time, reducing the need for complex external management systems while enhancing precision.
3Measurement precision
If optimization process matches historical recorded flight data with trajectory predictions by minimizing error functions, then prediction accuracy is enhanced, but loss of time increases due to continuous learning computations
Solution Approach 1:
The patent applies preliminary action by performing extensive optimization computations offline using historical data before actual prediction operations. The system pre-computes optimal parameter sets and stores them for rapid retrieval during real-time predictions, shifting the computational burden to offline processing and minimizing time loss during operational prediction phases.
Solution Approach 2:
The system applies partial action by selectively optimizing only the most critical trajectory parameters that have the greatest impact on prediction accuracy. Instead of continuously optimizing all possible parameters in real-time, the invention focuses computational resources on key parameters, achieving significant accuracy improvements with reduced computation time.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention proposes a new method to improve the accuracy of future predictions of arrival trajectories based on the use of historical flight recorded data of the traffic within a specific area approaching a designated airport. The invention is a method, which leverages the AIDL-based Trajectory Computation Infrastructure for improving the prediction capabilities of any Trajectory Predictor (TP) used for Arrival Management by means of a continuous learning process. This learning process determines what values of the selected parameters provide the most accurate trajectory predictions according to the aircraft type-airline-intended descent procedure-weather conditions combination.