Learning-Based User Association for Dual-UAV Enabled Wireless Networks With D2D Connections

When the communication infrastructures are damaged in disasters, unmanned aerial vehicles (UAVs) can be utilized as aerial base stations to achieve rapid service recovery. However, the wireless coverage of a single UAV is limited, and device-to-device (D2D) transmission can be exploited to accommodate more users with wireless service. Thus, in this paper, we consider the user association for a dual-UAV-enabled wireless network with the help of D2D connections in disasters. To achieve better performance, we maximize the weighted sum rate of the UAV-served users and the total number of D2D-connected users by optimizing the user association. The formulated problem is a combinatorial optimization problem involving binary variables, which is extremely difficult to solve. Accordingly, we propose two algorithms to solve it approximatively. The first algorithm is the learning-based clustering algorithm by viewing the optimization as a clustering problem. The users who can be served by the UAVs are regarded as cluster centers, which need to be selected optimally. In the second one, the binary variables are relaxed into continuous variables, and then, the problem can be solved by the existing optimization tools. The simulation results demonstrate that these two algorithms can achieve excellent suboptimal performance, and the computational complexity of the learning-based clustering algorithm is much lower.