Multi-Armed Bandit based Capacity Enhancement Approach for Reconfigurable Intelligent Surface Assisted Multi-cell System

In this paper, the capacity enhancement problem in multi-user networks is considered through the assistance of reconfigurable intelligent surfaces (RIS). RIS can manage the interference among cells by optimizing its phase and amplitude reflection coefficients, which is a non-convex problem. For an improved solution, we propose a Multi-armed Bandit (MAB) based technique to optimize the RIS phase shift for interference management while preserving users’ quality of service. In the proposed MAB, RIS phase shifts represent the arms, and the sum rate represents the corresponding reward. Two MAB algorithms are utilized to select the suitable arm from the available space, namely, the $\epsilon-$Greedy and the Decaying $\epsilon-$Greedy. The simulation results prove that the adopted RIS with phase shift optimization can significantly improve the rate for all users compared to the traditional network.