Machine Learning-Based Printed Lens Antenna With Graded-Index Metasurface for mmWave 5G NR N261 Mobile Communication Applications

This article presents an advanced metasurface-based printed antenna, leveraging a machine learning (ML) optimization method for millimeter-wave 5G New Radio (NR) communication networks. The ML-based Random Forest, Artificial Neural Networks (ANN), and XG Boost algorithms are developed and implemented to optimize the antenna performance parameters. The improvement of gain is discussed through the implementation of a phase-oriented graded-index meta-lens system for the designed 28-GHz 5G rectangular microstrip patch design (RMPD). The intended metamaterial (MTM) unit cells are designed and placed on the graded metasurface lens with a radial phase to analyze transmission characteristics. The designed meta-lens antenna offers gain enhancement by 2.32 dBi due to the meta-lens focusing effect in the intended direction. The designed patch antenna integrated with the lens exhibits a peak gain of around 8.33 dBi and efficiency above 98% within an operating band (27.5–29.1 GHz). The suggested metasurface antenna supports the 5G New Radio (5G NR) n261 (27.5–28.35 GHz) FR-2 band. The performance parameters of the simple patch antenna and metasurface Luneburg lens-integrated antenna structures have been evaluated and analyzed theoretically, followed by experimental validation of the fabricated prototypes. The finite element method-based full-wave simulation outcomes are validated with measurement results, which justify the correctness of the prescribed design approach to achieve improved gain for mm-wave 5G patch antenna.