Resource Allocation for Multi-Active-RIS assisted MISO Communications via Deep Reinforcement Learning

As the fifth-generation mobile communication system (5G) is commercially available, countries around the world have gradually launched research plans for the sixth-generation mobile communication system (6G). Reconfigurable intelligent surfaces (RISs), due to their advantages of easy deployment and low power consumption, have shown a potential application value in enhancing wireless transmission and coverage in 6G. In this paper, we consider a multiple active RISs-assisted downlink multiple-user multiple-input single-output (MISO) system, and study the problem of resource allocation. Specifically, we design a deep Q network (DQN) based resource allocation algorithm by regarding the active RISs as agent. By observing the real-time channel state information (CSI), the proposed algorithm can jointly optimize the user-RIS association matrix and the passive beam forming to maximize the system sum rate per unit bandwidth. Simulation results show that the proposed algorithm can approach the upper bound obtained by exhaustive search with relatively low time consumption.