Experimental Validation of Gaussian Process-Based Air-to-Ground Communication Quality Prediction in Urban...

Area of Science:

• Wireless Communication
• Machine Learning
• Channel Modeling

Background:

• Air-to-ground communication is crucial for urban relay missions.
• Accurate channel prediction is challenging due to complex urban environments.
• Existing methods often require extensive prior environmental information.

Purpose of the Study:

• To experimentally assess Gaussian Process (GP) regression for air-to-ground channel prediction.
• To develop and validate a method for simulating urban environments indoors.
• To compare GP-based prediction with empirical model-based prediction for relay missions.

Main Methods:

• Simulated an urban environment indoors using water containers to mimic buildings.
• Conducted indoor experiments to collect communication signal strength data.
• Applied Gaussian Process regression and empirical models for channel prediction.
• Evaluated prediction accuracy by comparing signal strength at optimal relay positions.

Main Results:

• Gaussian Process regression demonstrated superior performance compared to empirical models.
• GP-based prediction provided reasonable accuracy without needing a priori environmental data.
• The indoor simulation effectively replicated aspects of urban wireless propagation.

Conclusions:

• Gaussian Process regression is a viable and advantageous approach for dynamic urban air-to-ground channel prediction.
• The proposed indoor simulation method offers a practical alternative for experimental validation.
• GP regression's ability to work without prior environmental knowledge enhances its applicability in real-world scenarios.