Intelligent Reflecting Surface Assisted Over-the-Air Computation with Device-to-Device Communication

Over-the-air computation (OAC) enables devices to compute functions over distributed data without sharing local raw data and provides an efficient, distributed, and privacy-preserving computation mechanism. This is achieved by leveraging channel mixing properties and decoding results from received signals. However, optimizing the effective wireless channel between devices and the base station is critical for effective OAC. Intelligent reflecting surfaces (IRS) can enhance transmission quality, but the resulting optimization problem becomes complex with increasing the number of devices. This paper examines clustering-based OAC using an IRS in which users share their data with the cluster head before uplink transmission. The approach also helps addressing key scalability and efficiency challenges associated with OAC and intelligent surfaces. Simulations demonstrates that clustering reduces optimization complexity and improves convergence for alternating semidefinite relaxation (SDR) and difference-of-convex (DC) algorithms. Therefore, Clustering-based OAC provides an important step towards robust large-scale deployment.