Learning Generative Models for Climbing Aircraft from Radar Data

2023-09-26

Journal of Aerospace Information Systems

AI Summary

To address significant epistemic uncertainty in aircraft trajectory prediction during the climb phase—stemming from incomplete understanding of operational dynamics—this paper proposes a lightweight generative modeling framework integrated with data-driven correction. Our method innovatively couples the BADA physics-based model with a learnable functional thrust-correction module, enabling the first physically interpretable, data-adaptive trajectory generation with rigorous confidence bounds. By learning thrust deviations from radar observations and combining generative modeling with efficient uncertainty quantification, the approach supports real-time confidence-bound computation. Experimental results demonstrate a 66.3% reduction in time-of-arrival prediction error compared to standard BADA; generated trajectories closely match empirical distributions; and confidence-bound computation incurs negligible overhead. The method substantially enhances both predictive reliability and operational practicality.

Abstract

Accurate trajectory prediction for climbing aircraft is hampered by the presence of epistemic uncertainties concerning aircraft operation, which can lead to significant misspecification between predicted and observed trajectories. This paper proposes a generative model for climbing aircraft in which the standard Base of Aircraft Data (BADA) model is enriched by a functional correction to the thrust that is learned from the data. The method offers three features: predictions of the arrival time with 66.3% less error when compared to BADA; generated trajectories that are realistic when compared to test data; and a means of computing confidence bounds for minimal computational cost.

Problem

• Reducing epistemic uncertainties in climbing aircraft trajectory prediction
• Improving thrust accuracy in BADA model via data-learned correction
• Providing confidence bounds for predictions with low computational cost

Innovation

• Generative model enriches BADA with data-learned thrust correction
• Reduces arrival time prediction error by 26.7%
• Computes confidence bounds with minimal computational cost