Deep Reinforcement Learning for RIS-assisted Multi-UAV MU-MISO Communication Networks: Sum-rate and Energy Efficiency Maximization

Unmanned aerial vehicles (UAVs) have emerged as a promising solution for enhancing wireless networks, especially in challenging environments. However, recent studies that integrate reconfigurable intelligent surfaces (RIS) with UAVs tend to focus on limited aspects, such as single-UAV deployments or partial optimization of system parameters, thereby neglecting a comprehensive system-level design. To overcome these limitations, we propose a multi-user MISO communication network that leverages RIS-assisted UAVs to maximize both sum-rate and energy efficiency as two distinct objectives. Our approach stands out by considering multiple UAVs and incorporating four critical constraints: UAV flying areas, power limitations, transmit beamforming, and RIS requirements. We formulate separate optimization problems for sum-rate and energy efficiency, and address them using deep reinforcement learning (DRL) algorithms, namely proximal policy optimization (PPO) and deep deterministic policy gradient (DDPG). By jointly optimizing UAV coordinates, the transmit beamforming matrix, and RIS phase shifts, our method achieves significant performance improvements under dynamic environmental conditions. Extensive simulations show that our comprehensive strategy delivers higher sum-rates and enhanced energy efficiency, underscoring its practical potential for next-generation RIS-assisted UAV communication systems.