This paper proposes a reinforcement learning (RL)-based topology control system for unmanned aerial vehicle (UAV) networks. The system optimizes network connectivity by analyzing the relative positions of UAVs, considering interference and energy consumption, and reshaping the network's logical structure. Deep deterministic policy gradient (DDPG) is employed to allow each UAV to adapt its connectivity, minimizing learning time by adjusting the number of steps used for parameter learning. Simulations demonstrate the system's effectiveness in various topologies.

Taehoon Yoo, Sangmin Lee, Kyeonghyun Yoo, Hwangnam Kim