Stacked Intelligent Metasurfaces-Enabled Transceiver: Functional Coding and Data-Enhanced Deep Unfolding Detection

Recent studies have shown that the stacked intelligent metasurfaces (SIM) can exploit inter-layer electromagnetic (EM) wave transmission for wave-domain signal processing. This enables over-the-air computing with characteristics such as high-speed operation, low energy consumption, and support for multitasking parallel processing. By leveraging these advantages, SIM show great potential in substituting or augmenting customized communication functions in conventional wireless systems. To support various transmitter-customized communication functions, we design a novel multiple-input multiple-output (MIMO) transmitter architecture based on SIM and develop corresponding detection algorithms. Initially, we present a SIM-enabled MIMO transceiver model. Unlike conventional wireless communication, the SIM-enabled transmitter automatically performs customized functions, such as precoding and channel encoding, as the EM waves propagate within the layers. At the receiver side, minimum mean square error (MMSE) and maximum likelihood (ML) detectors are adopted as benchmarks, and their theoretical performance bounds are derived accordingly. Furthermore, we propose a data-enhanced deep unfolding algorithm, namely orthogonal approximate message passing causal dilated convolutional transformer (OAMP-CDT), to demonstrate the performance of customized functions based on SIM. Simulation results verify that the proposed SIM-enabled MIMO transmitter supports customized functions via over-the-air signal processing, and the proposed OAMP-CDT algorithm outperforms the same series of algorithms, such as orthogonal approximate message passing (OAMP) and OAMP-Net2.