Context-Aware Arrival Trajectory Prediction via Multiflow Informer with Environmental Influences

Abstract

Precise trajectory prediction in high-density terminal maneuvering areas is a fundamental prerequisite for the realization of next-generation trajectory-based operations. However, the practical deployment of deep learning models in this domain is often hindered by the technical challenges of effectively integrating heterogeneous environmental data and the inherent drift associated with recursive error accumulation. This study proposes a context-aware multiflow Informer framework that synergistically integrates target aircraft kinematics with operational traffic context and atmospheric perturbations. The architecture employs a cascaded gating mechanism to autonomously align internal flight dynamics with external influences, while utilizing a non-autoregressive generative decoder to achieve one-shot trajectory synthesis, thereby mitigating the cumulative error propagation characteristic of traditional recursive models. Experimental results using actual trajectory data from Guangzhou Baiyun International Airport demonstrate that the proposed model consistently outperforms recurrent baseline models, yielding architectural improvements ranging from 3.8 to 8.1% across multiple metrics and expanding to between 10.4 and 13.8% upon integrating multimodal environmental data. Furthermore, interpretability analysis suggests that the model has the potential to learn motion inertia and relevant operational logic, while cross-airport validation underscores its portability, offering a reliable predictive foundation for intelligent air traffic management systems.