Joint Deployment and Beamforming Design in IRS-Aided Networks With Discrete Location Space

In intelligent reflecting surface (IRS)-aided networks, in addition to base station (BS) beamforming and IRS phase shift, the location of the IRS also has a great impact on the communication rate. However, most works just consider location optimization in perfect continuous spaces, which will not work in more practical discrete spaces. This letter proposes an IRS-location-considered sum-rate maximization scheme, where a low-complexity IRS deployment algorithm with a discrete optimization space is designed, and BS beamforming and IRS phase shift are solved simultaneously instead of alternately. Specifically, we deduce a closed-form expression for the optimal IRS location in the single-user case and develop an efficient deployment procedure in the multi-user case by using the linear conic relaxation (LCR) to deal with the zero-one constraint. Furthermore, leveraging the second-order cone programming (SOCP), a novel parallel optimization algorithm for BS beamforming and IRS phase shift is designed, instead of traditional semi-definite relaxation (SDR) in an iterative manner. Finally, a sum-rate maximization problem with rate constraints is approximately solved by using the successive convex approximation and alternating optimization techniques. Numerical results show our proposed LCR-based deployment improves the sum-rate by 118.6% compared with the common centralized deployment. The SOCP-LCR-based scheme reduces the numbers of deployed IRSs and IRS elements by 60.1% and 29.8% while improving the sum-rate by 24.5% and 12.1% compared with the SOCP-based even deployment scheme and SDR-LCR-based scheme, respectively.