In this research, the performance of communication systems is studied an‎d enhanced through the use of Reconfigurable Intelligent Surfaces (RIS) in Communications Systems. These surfaces with their ability to control wave reflections via adjust phase shifts of the reflectors element on them, make it possible to improve channel conditions an‎d increase the capacity of communication systems. Given the high computational complexity an‎d the non-convex nature of many optimization problems in this field, Reinforcement Learning-based (RL-based) algorithms have been designed an‎d implemented in this thesis. These algorithms by interacting with the environment an‎d without requiring labeled data are able to effectively configure the elements phases of RIS. The objective of this research is to solve an optimization problem aimed at maximizing the capacity of a Single-User MIMO (SU-MIMO) systems using various reinforcement learning algorithms. Simulation results show that, under different channel an‎d system conditions, RL algorithms achieve performance nearly comparable to exact mathematical methods an‎d with only a slight reduction in efficiency, they can provide acceptable Performance. The main advantage of these algorithms lies in their significant reduction of computational complexity compared to classical methods an‎d their simpler feasibility for real-world implementation.

محمدجواد اميدي

حميد سعيدي سورك

فروغ السادات طباطباء , محمدرضا حيدرپور