Abstract
Reconfigurable Intelligent Surfaces (RIS) have emerged as a revolutionary technology for 6G wireless communication networks, offering the ability to programmatically control the indoor electromagnetic propagation environment. However, achieving optimal performance in indoor 6G scenarios requires solving highly non-convex joint optimization problems involving active beamforming at the access point (AP) and passive beamforming at the RIS, coupled with dynamic resource allocation under highly time-varying channel conditions. Traditional optimization-based approaches often struggle with computational complexity and fail to adapt to real-time changes. In this paper, we propose a novel deep reinforcement learning (DRL) framework based on the Deep Deterministic Policy Gradient (DDPG) algorithm to jointly optimize the active transmit beamforming at the AP, the passive phase shift matrix at the RIS, and the power allocation across multiple indoor user equipments (UEs). Our objective is to maximize the sum-rate of the system while satisfying the minimum quality-of-service (QoS) requirements for all users and complying with physical hardware constraints. Extensive simulation results demonstrate that the proposed DRL-based scheme converges rapidly and outperforms conventional optimization and baseline heuristic schemes in terms of spectral efficiency, robustness, and computational efficiency. This work provides an efficient, intelligent solution for real-time resource management in RIS-assisted indoor 6G environments.
