Convolutional Autoencoder-Based Phase Shift Feedback Compression for Intelligent Reflecting Surface-Assisted Wireless Systems

In recent years, intelligent reflecting surface (IRS) has emerged as a promising technology for 6G due to its potential/ability significantly enhance energy- and spectrum-efficiency. To this end, it is crucial adjust the phases of elements IRS, most research works focus on how optimize/quantize phase different optimization objectives. particular, quantized shift (QPS) assumed be available at which, however, does not always hold should fed back IRS in practice. Unfortunately, feedback channel generally bandwidth-limited, which cannot support huge amount overhead QPS particularly large number and/or quantization level each element. order break bottleneck, letter, we propose convolutional autoencoder-based scheme, compressed receiver side reconstructed side. solve problems mismatched distribution vanishing gradient, remove batch normalization (BN) layers introduce denosing module. By doing so, possible achieve high compression ratio with reliable reconstruction accuracy bandwidth-limited channel, also accommodate existing assuming IRS. Simulation results confirm feedback/compressed through show that proposed scheme can outperform algorithms.