Impact of Near-Field Conditions on the Performance of RIS-aided mmWave Communication Strategies

In this paper, we consider a reconfigurable intelligent surface (RIS)-based transmitter operating in a high-frequency band that sends more symbols per channel use than the number of antennas at the receiver. For such a system, we present a comprehensive study on the design and performance of a neural network (NN)-based receiver. We evaluate the performance of the minimum mean squared error (MMSE) receiver, which is a linear receiver, the optimum maximum likelihood (ML) detector, and the NN-based receiver for recovering the transmitted symbols. The numerical simulations carried out for all the detectors considered prove that the classical far-field simplification leads to a performance underestimation in scenarios involving wide transmitters and high-frequency bands, as the one considered in this paper. Moreover, the results show that the NN-based receiver achieves a favorable trade-off between complexity and performance, outperforming the MMSE receiver while maintaining a lower complexity than the ML detector.