An RIS-Assisted Integrated Deep Learning Framework for MIMO-OFDM-IM

Multiple-input multiple-output orthogonal frequency division multiplexing with index modulation (MIMO-OFDM-IM) offers a flexible trade-off between error performance and spectral efficiency. Meanwhile, reconfigurable intelligent surface (RIS)-assisted communication has emerged as a promising technology for optimizing wireless propagation environments, significantly enhancing signal quality and system performance. Integrating RIS into MIMO-OFDM-IM systems can improve communication reliability. However, efficient signal detection remains challenging due to severe interchannel interference and the correlation among subcarrier symbols within each subblock. To address this challenge, we propose InDeep, a robust deep learning (DL)-based detection framework designed for RIS-assisted MIMO-OFDM-IM systems. InDeep leverages one-dimensional convolutional neural network (1D-CNN) layers for feature extraction, followed by bidirectional gated recurrent unit (Bi-GRU) networks to capture temporal dependencies in the received signal. Additionally, we incorporate domain knowledge into the preprocessing of the channel matrix and received signal to enhance detection accuracy. Simulation results demonstrate that the proposed RIS-assisted MIMO-OFDM-IM system with InDeep detection significantly improves communication reliability, making it a compelling solution for next-generation wireless networks.