Dynamic Resource Allocation for MmWave UAV Communications: A Deep Reinforcement Learning Approach

Millimeter wave (mmWave) enabled unmanned aerial vehicle (UAV) communications featured by high flexibility and data rate, are widely regarded as an essential element of 6G networks. This paper focuses on the dynamic resource allocation of mmWave UAV communication systems. This problem as a joint optimization of the 3D UAV trajectory, beamwidth and power allocation, with the objective of maximizing normalized spectral efficiency (NSE). Considering that this problem is non-convex and can not be solved directly with the traditional methods, we propose to decouple it into two tractable sub-problems. Moreover, we propose two deep deterministic policy gradient (DDPG)-based algorithms to effectively find the optimal solution in continuous space. Simulation results show that the proposed DDPG-based algorithms can significantly improve throughput.