Public defence in Communications Engineering and Networking Technology, M.Sc. Yu Bai

The title of the thesis: Deep Reinforcement Learning-Driven Optimization for UAV-Enabled Wireless Networks

Thesis defender: Yu Bai
Opponent: Prof. Hirley Alves, University of Oulu, Finland
Custos: Prof. Riku Jäntti, Aalto University School of Electrical Engineering

Unmanned aerial vehicles (UAVs), commonly known as drones, are increasingly used to provide wireless connectivity in situations where conventional infrastructure is limited or disrupted, such as disaster areas, rural regions, or large-scale events. However, managing fleets of UAVs for communication and sensing requires fast and intelligent decision-making about where drones should fly, how they should allocate their communication resources, and how they should coordinate with ground users. 

This doctoral research explores how deep reinforcement learning (DRL), which is a branch of artificial intelligence that enables systems to learn from interaction and feedback, can be applied to optimize UAV-enabled wireless networks. The study developed new learning-based frameworks that allow UAVs to make adaptive and real-time decisions in three important scenarios: 

Dynamic multi-UAV deployment: UAVs acting as flying base stations that automatically adjust their positions, modes, and transmission power to balance energy use and network coverage. 

Data collection with movable antennas: UAVs equipped with steerable antennas that shorten flight paths and improve data collection efficiency in wireless sensor networks. 

Integrated sensing and communication (ISAC): UAVs that jointly perform radar sensing and data transmission while minimizing the age of information, a metric that reflects how fresh and timely the delivered information is—crucial in tasks like disaster relief and transportation safety. 

The results demonstrate that DRL-based methods can substantially improve energy efficiency, adaptability, and timeliness compared to traditional optimization strategies. The thesis introduces new system models, algorithms, and evaluation frameworks that highlight the potential of intelligent UAV networks to become more autonomous and resilient. 

These findings contribute to the development of future 6G communication systems, offering practical insights for applications ranging from emergency communications to the Internet of Things and intelligent transportation. The research paves the way for more reliable and efficient aerial communication infrastructures in real-world environments.

Key words: Unmanned Aerial Vehicles (UAVs), Wireless Communication, Deep Reinforcement Learning

Thesis available for public display 7 days prior to the defence at Aaltodoc.