Machine Learning-Based NOMA in LEO Satellite Communication Systems

In nonorthogonal multiple access (NOMA), which is one of the technologies attracting attention due to its high frequency efficiency in wireless communication systems, the decoding order is considered a crucial factor influencing system performance. However, finding the optimal decoding order becomes challenging when there is wide service coverage and multiple users due to the increased computational complexity involved in the process. In this paper, we propose a machine learning (ML)-based NOMA in low earth orbit (LEO) satellite communication systems. Our proposed scheme aims to utilize ML to determine the optimal decoding order that satisfies the target signal-to-interference-plus-noise ratio (SINR) constraints with low computational complexity, particularly in systems where the satellite utilizes beamforming vector and NOMA to serve each user. The numerical results demonstrate that our proposed scheme achieves performance comparable to that of the optimal scheme for obtaining the optimal decoding order by using the iterative calculation.