Machine Learning Approaches for Intelligent Reflecting Surface-Assisted MIMO Communication Systems: A Survey

The intelligent reflecting surface (IRS) is an emerging research concept that enables programmable wireless propagation. Multiple-input multiple-output (MIMO) is one of the most widely used communication systems. To enhance user performance, machine learning (ML) provides a variety of techniques to improve signal quality. This survey systematically reviews recent advances in ML-assisted IRS-MIMO to identify challenges and research opportunities. A total of ten research papers and two survey papers were analyzed. The reviewed studies are categorized according to their ML models and their specific applications in channel estimation (CE), pilot overhead reduction, bit error rate (BER)/symbol error rate (SER) minimization, and the maximization of sum rate, throughput, and data rate. The study presents comparisons and highlights the contributions and limitations of existing methods. Finally, the survey outlines future directions for the practical implementation of ML-assisted IRS-MIMO systems.