A digital twin-driven computation and analysis framework for low-altitude airspace.

Area of Science:

• Aerospace Engineering
• Computer Science
• Artificial Intelligence

Background:

• Low-altitude airspace management faces challenges like high dynamic complexity, safety risks, and fragmented information.
• Existing systems struggle with real-time analysis and prediction in dynamic environments.

Purpose of the Study:

• To propose a digital twin-enabled computation and analysis framework for low-altitude airspace management.
• To enhance spatial mapping, spatiotemporal representation, and dynamic modeling.
• To improve trajectory prediction accuracy and conflict warning capabilities.

Main Methods:

• A four-layer digital twin architecture integrating multi-source data fusion.
• Implementation of a bidirectional GRU-Seq2Seq trajectory prediction model.
• Inclusion of a Kalman filter for error compensation and a bidirectional physical-virtual closed-loop interaction.

Main Results:

• Achieved an average trajectory prediction error of 1.52 m, outperforming baseline models by up to 47.9%.
• Maintained low prediction error (max 3.9 m at 15s) across varying prediction horizons.
• Demonstrated superior robustness in GPS denial and communication interruption scenarios (avg. error 2.15 m and 2.38 m).

Conclusions:

• The digital twin framework significantly enhances low-altitude airspace management performance.
• The proposed method outperforms traditional geometric twins and deep learning approaches in accuracy, stability, and robustness.
• The framework is validated as effective and practical for real-world low-altitude airspace operations.