This paper investigates an Intelligent Reconfigurable Surface (IRS) enhanced multi-user uplink Multiple-Input-Single-Output (MISO) communication system consists of an Access Point (AP), an IRS and several user devices. We assume that the activation IRS reflecting elements number is limit due to the consideration of the energy efficiency on IRS. To maximize the sum of uploading transmission rate of all the user devices, phase shift and elements selection are jointly optimized. The formulated optimization problem involves integer and continuous optimization variables, requiring high complexity conventional optimization-based approaches. Provided that the computational bottleneck lies in dealing with integer variables, we propose a Deep Reinforcement Learning (DRL) based approach, in which the integer variables (activation IRS reflecting elements selection pattern) are determined by the output of a neural network and the remaining continuous variable (phase shits of the corresponding activation elements) are determined by using an existing conventional algorithm. Moreover, a novel network architecture is developed to construct the policy and Q network in the proposed DRL-based approach. Numerical results indicate the effectiveness of the proposed DRL based approach.

A Deep Reinforcement Learning Based Approach for Intelligent Reconfigurable Surface Elements Selection

Gravina R.;
2022-01-01

Abstract

This paper investigates an Intelligent Reconfigurable Surface (IRS) enhanced multi-user uplink Multiple-Input-Single-Output (MISO) communication system consists of an Access Point (AP), an IRS and several user devices. We assume that the activation IRS reflecting elements number is limit due to the consideration of the energy efficiency on IRS. To maximize the sum of uploading transmission rate of all the user devices, phase shift and elements selection are jointly optimized. The formulated optimization problem involves integer and continuous optimization variables, requiring high complexity conventional optimization-based approaches. Provided that the computational bottleneck lies in dealing with integer variables, we propose a Deep Reinforcement Learning (DRL) based approach, in which the integer variables (activation IRS reflecting elements selection pattern) are determined by the output of a neural network and the remaining continuous variable (phase shits of the corresponding activation elements) are determined by using an existing conventional algorithm. Moreover, a novel network architecture is developed to construct the policy and Q network in the proposed DRL-based approach. Numerical results indicate the effectiveness of the proposed DRL based approach.
2022
978-1-6654-6297-6
activation IRS reflecting elements selection
deep reinforcement learning
intelligent reconfigurable surface
phase shift design
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.11770/360954
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 2
  • ???jsp.display-item.citation.isi??? 0
social impact