Transceiver Design for IRS-Aided Massive MIMO Networks With Rate-Splitting

Intelligent reflecting surface (IRS), equipped with multiple reflective elements, is an important technology for improving the achievable rate and energy efficiency of communications. Meanwhile, rate-splitting (RS) is an approach that reduces the detrimental effect of multi-user interference in the system by splitting users’ messages into private and common parts. In this work, we use the RS method in an IRS-aided massive multiple-input multiple-output (mMIMO) system. We assume each user is equipped with multiple antennas and then design the transceiver, which include the precoder and the combiner. We analyze the performance of the achievable sum-rate when the base station (BS) and users know perfect channel state information (CSI) or only statistical CSI. We use a projected gradient descent method (PGDM) to optimize the IRS phase shifts by minimizing the mean-squared error (MSE) for data estimates when the system has mixed CSI. The performance of combiner and precoder quantization, as well as the limited feedback has also been investigated in simulations in terms of the achievable sum-rate by using the iterative chordal distance (ICD) method.