An Optimal Reflective Elements Grouping Model for RIS-Assisted IoT Networks Using Q-Learning

Abstract

One of the critical challenges of the next-generation Internet of Things (IoT) network is to ensure reliable and low latency data transfer at the IoT devices (IoD), while minimizing the overall power consumption. Data latency and throughput performance degrade in a dense IoT network due to obstacles between IoD and the access point (AP). Therefore, an efficient signal processing model needs to be developed to address the aforementioned challenges. Towards this end, in this work, an optimal reflective elements (RE) grouping model over reconfigurable intelligent surface (RIS)-assisted IoT network is proposed. The proposed RE grouping model enhances the signal strength at the IoD when the direct link between AP and IoD is blocked. Additionally, in case of reliable direct link between AP and IoD, reflected beam power is controlled over RIS-IoD link, resulting in minimized power consumption. Hence, in this work, a novel method to regularize the power consumption over the IoT network is proposed. The proposed method groups the RE to reflect the incident signal. This grouping is based on the joint common group phase-shift and channel correlation between RE. Moreover, an optimization problem is formulated to minimize the total latency with respect to the received power and RIS reflection matrix. Due to the coupled nature of constraints, the optimization problem is non-convex and is solved using the Q-learning. Finally, the performance of the proposed method is compared with the benchmark methods, taking various network parameters into account.