Optimizing Over-the-Air Computation in IRS-Aided C-RAN Systems

Abstract

Over-the-air computation (AirComp) is an efficient solution to enable federated learning on wireless channels. Air-Comp assumes that the wireless channels from different devices can be controlled, e.g., via transmitter-side phase compensation, in order to ensure coherent on-air combining. Intelligent reflecting surfaces (IRSs) can provide an alternative, or additional, means of controlling channel propagation conditions. This work studies the advantages of deploying IRSs for AirComp systems in a large-scale cloud radio access network (C-RAN). In this system, worker devices upload locally updated models to a parameter server (PS) through distributed access points (APs) that communicate with the PS on finite-capacity fronthaul links. The problem of jointly optimizing the IRSs' reflecting phases and a linear detector at the PS is tackled with the goal of minimizing the mean squared error (MSE) of a parameter estimated at the PS. Numerical results validate the advantages of deploying IRSs with optimized phases for AirComp in C-RAN systems.

Wireless Network Coding With Intelligent Reflecting Surfaces

“…Hence, the desired realization of the channels can be achieved, meaning that the interference issue and the wireless channel fading impairment can be tackled fundamentally. Integrated electronics can be used to control IRS [11,13,14].…”

“…Due to the above benefits, IRSs can enhance wireless communication performance in terms of coverage, energy efficiency, electromagnetic radiation reduction, and wireless localization accuracy [16][17][18][19][20]. Also, IRSs can improve the BER performance of over-air-computation (OAC) techniques, which leverage the superposition property in multiple-access channel from sources to a relay [14,21,22]. It can achieved by intelligently tuning the IRS phase profile, thereby significantly boosting the received signal power at the desired receiver.…”

Federated learning and next generation wireless communications: A survey on bidirectional relationship

“…Their method's objective was to simultaneously optimize device selection, IRS's phase shifts and the aggregation beamformer so as to cancel out the higher model aggregation errors when the number of client devices are maximized and the optimization problem was solved using a difference of convex (DC) algorithm. Furthermore, Yu et al [68] discussed the benefits of IRS assisted AirComp in a cloud radio access network. The system model of [68] consisted of distributed access points (APs) to which the local model updates were sent and each AP forwarded the received update signals to the PS through the finite capacity fronthaul link.…”

“…Furthermore, Yu et al [68] discussed the benefits of IRS assisted AirComp in a cloud radio access network. The system model of [68] consisted of distributed access points (APs) to which the local model updates were sent and each AP forwarded the received update signals to the PS through the finite capacity fronthaul link. The authors designed an iterative algorithm for optimizing the reflecting phases of the IRS along with linear detection vector of the global PS.…”

Learning Rate Optimization for Federated Learning Exploiting Over-the-Air Computation

“…Meanwhile in [27], joint device selection and receive beamforming design was investigated in single-input-multiple-output (SIMO) configuration, and a novel unified difference-of-convex (DC) function was proposed. On the other hand, the problem of distortion minimization in an intelligent reflection surface (IRS)-aided cloud radio access network (CRAN) system was addressed in [28], where a joint optimization scheme of the passive beamforming and linear detection vector was designed. Furthermore, with the aid of multi-IRS, a novel resource and device selection method was developed to minimize the aggregate error as well as maximize the selected devices [29].…”