Machine Learning for the Dynamic Positioning of UAVs for Extended Connectivity.

Area of Science:

• Computer Science
• Electrical Engineering
• Telecommunications

Background:

• Unmanned Aerial Vehicle (UAV) networks offer versatile solutions for cellular network reinforcement.
• UAVs can address user demands exceeding ground base station capacity or compensate for base station failures.

Purpose of the Study:

• To develop and evaluate a UAV positioning algorithm for balanced traffic redistribution.
• To minimize network congestion and reduce the number of users without a connection using UAVs.

Main Methods:

• Utilizing machine learning algorithms to predict overloaded traffic areas.
• Implementing a dynamic UAV positioning algorithm to optimize traffic flow.
• Testing the algorithm with real-world user connection data.

Main Results:

• The UAV positioning algorithm significantly reduces unconnected users, particularly under less restrictive network conditions.
• Algorithm performance is notably better in less restrictive network scenarios compared to more restrictive ones.
• Prediction accuracy is crucial for both reducing unconnected users and optimizing UAV deployment.

Conclusions:

• Dynamic UAV placement effectively enhances cellular network performance by balancing traffic.
• Accurate traffic prediction is essential for maximizing the benefits of UAV-assisted networks.
• The proposed algorithm demonstrates significant potential for improving mobile network reliability and user experience.