Resource Allocation Schemes for Scalable Panel-Based LIS Surfaces

Panel-based large intelligent surface (LIS) systems, where each panel comprises a lower, but still significant, number of antennas, and is equipped with a number of baseband outputs (generally much smaller than the number of terminals), allow for a direct dimensionality reduction. However, such a dimensionality reduction may not be enough to simplify the overall system. This paper focuses on fully decentralized system architectures that allow for promising performances while requiring low, or even no, central data processing. In particular, this paper proposes panel-selection and panel-terminal association algorithms for decentralized panel-based LIS communication systems aiming at allocating a set of terminals to a given panel, which is limited to a given number of outputs, in a sequential or flooding manner, while maximising the minimum terminal rate. Performance results show that sequential and flooding distributed system architectures offer decent convergence rates in comparison with a centralized system architectures when selecting a proper combination of system parameters (i.e., number of panels, number of terminals, and number of baseband outputs), in turn ensuring scalability. In particular, it is shown that, for a given combination of the system parameters, a max-min terminal rate achieved when considering sequential and flooding distributed system architectures can successfully converge to the one achieved by the centralized system architecture, occurring after just a few iterations. Both the proposed sequential and flooding algorithms represents a less complex, but still an effective way to perform both panel selection and panel-terminal association, allowing for the deactivation of a significant portion of panels without substantially degrading the max-min terminal rate achieved by the centralized system architecture.