Statistical analysis and prediction of dynamic UAV-based entanglement distribution channel through deep learning

Area of Science:

• Quantum Information Science
• Artificial Intelligence
• Aerospace Engineering

Background:

• Unmanned aerial vehicles (UAVs) offer flexible quantum communication but face challenges from dynamic atmospheric channels.
• Traditional statistical models fail to capture real-time atmospheric variations, limiting UAV-based entanglement distribution performance.

Purpose of the Study:

• To develop an AI-driven adaptive prediction framework for forecasting UAV-to-ground quantum channel transmittance.
• To enhance the reliability and optimization of UAV-based quantum entanglement distribution systems.

Main Methods:

• Proposed an AI framework utilizing the Patch Time Series Transformer (PatchTST) model.
• Simulated UAV-to-ground quantum channel transmittance.
• Compared PatchTST performance against standard recurrent neural network benchmarks.

Main Results:

• The PatchTST model accurately predicted dynamic transmittance fluctuations in the UAV quantum channel.
• The proposed method outperformed recurrent neural networks in capturing essential channel features.
• Enabled reliable estimation of key performance metrics, such as the BBM92 secure key rate.

Conclusions:

• An effective deep learning strategy for real-time channel prediction in UAV-based quantum communication was demonstrated.
• The framework enhances adaptive optimization and robustness for UAV-based quantum entanglement distribution.
• This approach addresses limitations of static models in dynamic atmospheric conditions.