SE-Based User Clustering for Sum Rate Maximization in Multi-UAV NOMA Networks

This paper proposes a user clustering algorithm for a non-orthogonal multiple access (NOMA) network where multi-UAVs serve as base stations for multiple ground users. The joint optimization of the user clustering and power allocation is required to maximize the sum rate. However, this is an intractable problem in which two problems are coupled together. Although the user clustering by the brute-force search can obtain the optimal solution, the computational complexity is very high. For the low-complexity approach, we propose a squeeze-and-excitation (SE)-based user clustering algorithm. In our method, the model which is trained to maximize the sum rate is used to find the user clusters. Numerical results show that the SE-based algorithm outperforms the conventional user clustering algorithm by exploiting high representational power of the SE blocks. The results also show that our SE-based algorithm has a much lower complexity than the brute-force method.