Two-Timescale Optimization for Intelligent Reflecting Surface-Assisted MIMO Transmission in Fast-Changing Channels

The application of intelligent reflecting surface (IRS) depends on the knowledge channel state information (CSI), and has been hindered by heavy overhead training, estimation, feedback in fast-changing channels. This paper presents a new two-timescale beamforming approach to maximizing average achievable rate (AAR) IRS-assisted MIMO systems, where IRS is configured relatively infrequently based statistical CSI (S-CSI) base station precoder power allocation are updated frequently quickly outdated instantaneous (I-CSI). key idea that we first reveal optimal small-timescale I-CSI yields water-filling structure. Given allocation, mini-batch sampling (mbs)-based particle swarm optimization (PSO) algorithm developed optimize large-timescale configuration with reduced samples. Another important aspect develop model-driven PSO configuration, which maximizes lower bound AAR only using S-CSI eliminates need serves as dependable for mbs-PSO. Simulations corroborate superiority strategy its alternatives terms efficiency, benefits demonstrated.