DQN-Based Routing Resources Optimization in UAV Swarm Communication System

Unmanned Aerial Vehicles (UAVs) can collect a wide variety of data in the air and then transmit them to the ground stations, because of the limited capability of a single UAV, the application of UAV swarm communication system is becoming more and more widespread. Two data communication modes are included, the UAV-to-base station (U2B) link and the UAV-to-UAV (U2U) link. According to the principles of deep Q learning networks(DQN), treated as an agent the U2U link could interact with the UAV swarm communication system to select appropriate routing nodes for data transmission and channel attention(CA) is introduced to further improve system performance. The simulation data shows that the network based on reinforcement learning can obtain higher data transmission capacity with fewer routing hops.