Reinforcement Learning-based Energy Efficiency Optimization for RIS-Assisted UAV Hybrid Uplink and Downlink System

As Reconfigurable Intelligent Surface (RIS) technology can intelligently reflect the incident signal, it can be deployed to maintain a line-of-sight connection and improve the signal strength between unmanned Aerial Vehicles (UAV) and users. Current researches on RIS-assisted UAV communication typically consider optimizing data rates for single-line communications and do not focus on the energy consumption of the UAV. Furthermore, it is frequently challenging to apply conventional algorithms to RIS-assisted UAV optimization problems in a timely way, and the two-dimensional trajectory design cannot fully utilize the 3D mobility of the UAV. In order to increase the data rates for all ground users while improving the energy efficiency of the system, we consider a RIS-assisted full-duplex UAV communication system, in which a DDQN-based algorithm is proposed to jointly optimize the phase shift of the RIS and the 3D trajectory of the UAV. Simulation results demonstrate that the proposed algorithm can achieve a significant data rate gain and energy efficiency compared to time-division duplex systems and other deep reinforcement learning algorithms.