Multigraph spectral clustering for joint content delivery and scheduling in beam-free satellite communications

This paper tackles the problem of user scheduling in satellite content delivery networks with precoding. The clustering process has to consider two crucial and independent characteristics of the user terminals. On the one hand, users belonging to the same group shall have a reduced Euclidean norm between their channel vectors in order to obtain the maximum precoding gain. On the other hand, with the aim of exploiting the multicast capabilities of the system, user terminals grouped in the same cluster shall have requested the same content. The resulting clustering problem is formulated as a multigraph (also known as multiview) spectral clustering problem. The paper shows that this unsupervised learning framework is able to capture the different peculiarities of the mentioned problem. Two different techniques are introduced and validated in a close-to-real numerical simulation.
This work has received funding from the Spanish ministry of Science, Innovation and Universities under project TERESA-TEC2017-90093-C3-1-R (AEI/FEDER,UE), and by the Catalan Government under grant 2017-SGR01479.
Peer Reviewed
Postprint (author's final draft)