Machine Learning Driven Design and Optimisation of a Dual-Band SRR Metamaterial Antenna for Emerging IoT Platforms

Abstract

This research introduces a machine learning (ML) technique derived from the dual-band split-ring resonator (SRR) metamaterial antenna development and optimisation. The patch antenna with an SRR design is intended for use in 5G and 6G IoT applications. The method utilizes data driven prediction models to speed up design and optimisation instead of a normal trial and error process. The antenna responses were anticipated using different regression methods, including artificial neural network (ANN), K nearest neighbour (KNN), gradient boosting (GB), and eXtreme gradient boosting (XGBoost). When it comes to the techniques utilized, the eXtreme gradient boosting model was the one that reached the highest accuracy of 87.4 % in estimating return loss and thus it was the one that stood out the most. To support the claim of the effectiveness of the method, a dual-band split-ring resonator loaded microstrip antenna was designed and produced to operate in the ranges of 2.3 GHz to 2.55 GHz and 3.95 GHz to 4.15 GHz, respectively. The close agreement between predictions, simulations, and measurements further confirms the reliability of the machine learning-based design strategy. The proposed approach not only enables fast prototyping of compact antennas suitable for IoT devices, but also for industrial automation, smart home, and smart city applications.