Low-latency Data Transfer over SNR Constrained IoT Networks Using Optimal Active RIS Elements: A Q-learning Framework

Abstract

Reconfigurable Intelligent Surface (RIS) with active reflecting elements (REs) draws great attention due to its ability to improve the performance of the Internet of Things (IoT) networks. However, when used in its inherently passive nature, RIS-assisted networks suffer from the multiplicative loss in the reflected path between the access point (AP) and IoT devices (IoDs). Active RIS has the potential to amplify and reflect the incident signal towards the receiving end, alleviating the path loss. This paper considers a multi-RIS-assisted IoT network where multiple antenna-equipped AP serves the distant single antenna IoDs via orthogonal beams. These beams are amplified and reflected through multiple RISs. Consequently, multi-hop cascaded links between AP and IoDs are formed with the help of selected RISs. The proposed routing protocol enables the standards and semantics of how the signal will propagate throughout the network resulting in reduced data latency. In order to address the aforementioned challenges, this paper proposes a Q-learning framework for efficient routing over a RIS-assisted IoT network. An upper bound on the optimal number of active REs is also computed herein. Furthermore, an optimization problem is formulated to minimize the data latency with respect to active RIS reflection properties, minimum allowable signal-to-noise ratio (SNR) of the cascaded link, and RIS power consumption. Finally, the performance of the proposed method is compared with the existing methods, taking into account various network parameters such as the optimum number of active REs needed to optimize the network SNR and data latency.