Latency-Aware Computation Offloading in Multi-RIS-Assisted Edge Networks

Abstract

Mobile Edge Computing (MEC) has a widely established merit of bringing powerful computing servers to geographically closer locations to computationally limited devices; hence, reducing the task offloading latency from these devices to the servers. However, the frequent communication between devices and edge servers increases the network-wide traffic, therefore, stands in the way of enjoying notable improvements in network-wide latency. The optimization of this problem becomes of utmost importance to boldly underline the advantages of MEC-based task offloading. Reconfigurable Intelligent Surfaces (RISs) offer the potential to improve wireless transmission environment, thereby contributing to achieving this objective. This paper addresses a network-wide latency optimization problem in the context of multi-RIS-assisted MEC offloading scenarios. This problem is, herein, subdivided into two sub-problems, namely: ${a}$ ) path selection and ${b}$ ) joint optimization of RIS phases and offloading volume. The cascaded channel gain is derived for ${a}$ ). A balanced trade-off is sought between passive beamforming gain and reflection loss by reformulating ${a}$ ) as a shortest path problem using Graph Theory (GT). ${b}$ ) requires a different approach than the typical Successive Convex Approximation (SCA) technique used for single RIS. Instead, a Semi-Definite Relaxation (SDR) method is adopted to transform the non-convex sub-problem into a convex one, solved efficiently using an Alternating Optimization (AO) approach. The results of extensive simulations and numerical analyses reported herein reveal the notable latency reduction potency of multi-RIS-assisted MEC systems by, respectively, 24% as compared to their RIS-agnostic counterparts, and, by 15.34% when compared to systems incorporating a single RIS.

Keywords

• Reconfigurable intelligent surface (RIS)
• mobile edge computing
• latency minimization
• alternating optimization