Deep Learning Based Antenna Selection Technique for RIS-Empowered RQSM System

Abstract

Reconfigurable intelligent surface (RIS) technology has attracted considerable interest due to its ability to control wireless propagation with minimal power usage. Receive quadrature spatial modulation (RQSM) scheme transmits data bits in both in-phase (I) and quadrature (Q) channels, doubling the number of active receive antenna indices and improving spectral efficiency compared to the traditional receive spatial modulation (RSM) technique. Also, capacity-optimized antenna selection (COAS) improves error performance by selecting antennas with the best channel conditions. This paper proposes a new deep neural network (DNN)-based antenna selection method, supported by the COAS technique, to improve the error performance of the RIS-RQSM system. Monte Carlo simulations of the proposed DNN-COAS-RIS-RQSM system using the quadrature amplitude modulation (QAM) technique for Rayleigh fading channels are performed and compared with the COAS-RIS-RQSM system. Also, a comparative analysis of the computational complexities of the DNN and COAS techniques is conducted to evaluate the trade-offs between error performance and complexity.