Informative Path Planning Using Physics-Informed Gaussian Processes for Aerial Mapping of 5G Networks.

Area of Science:

• Telecommunications Engineering
• Robotics and Control Systems
• Statistical Machine Learning

Background:

• 5G private cellular networks are increasingly used in industrial settings.
• Ensuring reliable coverage and boundary requirements is critical but measurement-intensive.
• Unmanned Aerial Vehicles (UAVs) offer a potential solution for efficient network mapping.

Purpose of the Study:

• To propose a method using UAVs and Gaussian Processes (GPs) for efficient 5G private network coverage mapping.
• To enhance GP prediction accuracy through physics-informed mean functions and ray-tracing simulations.
• To develop an informative path-planning algorithm for optimized UAV-based network surveying.

Main Methods:

• Integration of physics-informed mean functions (including ray-tracing) into GP models.
• Quantitative evaluation of different GP mean functions using real-world 5G RSRP measurements.
• Development and hardware-in-the-loop simulation of an informative path-planning algorithm combining GPs and Bayesian optimization.

Main Results:

• Appropriate mean functions significantly enhance GP prediction accuracy.
• The informative path-planning algorithm reduces UAV flight distance by up to 98%.
• Maintained an average root-mean-square error of less than 6 dBm compared to measurement trials.

Conclusions:

• UAVs combined with physics-informed GPs provide an accurate and efficient solution for 5G private network mapping.
• The proposed informative path-planning significantly optimizes data collection efforts.
• This methodology is crucial for reliable industrial 5G deployment and coverage validation.