Metasurface Absorber-Integrated MIMO Antenna for Sub-6 GHz 5G Network with Autoencoder-Assisted Absorptivity Prediction
摘要
This work proposes an autoencoder-based deep learning optimization technique for the prediction and design of a metasurface absorber to attain the maximum absorption characteristics. A two-port and 4-port MIMO antenna has been designed and investigated for 5G sub-6 GHz applications. The isolation between the MIMO antenna units has been increased by the incorporation of a metasurface absorber between the antenna unit elements. The measured impedance bandwidth of the proposed metasurface absorber integrated antenna module is observed of 4.9–5.76 GHz (860 MHz) with isolation of more than 23 dB. Further, this work presents a decoder-free autoencoder–regressor framework for supervised regression, leveraging nonlinear latent features to enhance absorptivity prediction. Diagnostic analyses confirm that latent embeddings reduce redundancy, improve generalization, and achieve efficient, accurate performance. The results indicate that MIMO antenna components may be appropriately isolated while maintaining a planar geometrical layout. The proposed metasurface absorber-based MIMO antenna has an envelope correlation coefficient (ECC) under 0.01, a diversity gain (DG) approaching 10 dB, and MIMO component isolation over 23 dB.