Joint Active and Passive Beamforming in RIS-Assisted Covert Symbiotic Radio Based on Deep Unfolding

• Published in: IEEE transactions on vehicular technology Vol. 73; no. 9; pp. 14021 - 14026
• Main Authors: He, Xiuli, Xu, Hongbo, Wang, Ji, Xie, Wenwu, Li, Xingwang, Nallanathan, Arumugam
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.09.2024; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Array signal processing; Artificial neural networks; Backscatter; Beamforming; Communications systems; Constraints; Covert communication; deep unfolding; Design optimization; Multilayers; Optimization; Radio; Receivers & amplifiers; reconfigurable intelligent surface; Reconfigurable intelligent surfaces; Signal to noise ratio; Symbiosis; symbiotic radio; Symbols
• ISSN: 0018-9545; 1939-9359
• DOI: 10.1109/TVT.2024.3393724

In this paper, we consider an reconfigurable intelligent surface (RIS)-assisted multiple-input-single-output (MISO) covert symbiotic radio (SR) communication system. RIS, as a secondary transmitter (STx), can enhance primary transmission from the primary transmitter (PTx) to the primary receiver (PRx). Simultaneously, STx transmits its own information to the secondary receiver (SRx). In addition, RIS-assisted covert communications have a broad development prospect, and covert communication is considered in which Willie eavesdrops passive signals from RIS (Alice) to SRx (Bob). By jointly optimizing active beamforming vector at PTx and passive beamforming matrix at RIS, the achievable rate of PRx is maximized subject to the covertness constraint and the signal-to-noise ratio (SNR) constraint for secondary transmission. The optimization problem is challenging because of the non-convex objective function and the coupling between variables. Thus, the deep unfolding algorithm based on gradient descent (DUAGD) is proposed for the beamforming design. Specifically, we first transform the optimization problem with constraints into the dual domain. Then inspired by gradient descent algorithm, deep unfolding unfolds the original iterative process into a multi-layer network structure. Results from simulations show that the proposed algorithm has fast convergence while maintaining performance.