Flight Delay Prediction via Cross-Modality Adaptation of Large Language Models and Aircraft Trajectory Representation

2025-10-24

AI Summary

To address low prediction accuracy for terminal-area flight delays and the difficulty of fusing heterogeneous multi-source information, this paper proposes a lightweight multimodal forecasting framework. We introduce a novel “trajectory-to-language” cross-modal adaptation mechanism that encodes structured flight trajectories into text-like sequences, enabling unified input—alongside aviation weather reports and operational instructions—into a large language model (LLM). Through joint representation learning and dynamic context modeling, the framework achieves fine-grained causal attribution and real-time delay inference. Experiments demonstrate an average prediction error of under 3.2 minutes in minute-level forecasting—outperforming baseline models by 27.6%—while supporting low-latency online updates and practical deployment. The core contribution lies in bridging the modality gap between trajectory data and LLMs, establishing an interpretable and scalable paradigm for intelligent air traffic scheduling.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Large language models for search
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Flight delay prediction has become a key focus in air traffic management, as delays highlight inefficiencies that impact overall network performance. This paper presents a lightweight large language model-based multimodal flight delay prediction, formulated from the perspective of air traffic controllers monitoring aircraft delay after entering the terminal area. The approach integrates trajectory representations with textual aeronautical information, including flight information, weather reports, and aerodrome notices, by adapting trajectory data into the language modality to capture airspace conditions. Experimental results show that the model consistently achieves sub-minute prediction error by effectively leveraging contextual information related to the sources of delay. The framework demonstrates that linguistic understanding, when combined with cross-modality adaptation of trajectory information, enhances delay prediction. Moreover, the approach shows practicality and scalability for real-world operations, supporting real-time updates that refine predictions upon receiving new operational information.

Problem

• Predicting flight delays using multimodal data integration
• Adapting trajectory data into language modality for analysis
• Enhancing delay prediction through cross-modality information fusion

Innovation

• Adapts large language models for cross-modality prediction
• Integrates trajectory data with textual aeronautical information
• Achieves sub-minute error via contextual delay source analysis