Machine Learning based Low-Scale Dipole Antenna Optimization using Bootstrap Aggregation

Abstract

Dipole antennae are commonly used radio frequency devices. They gained good prominence as a result of their efficiency, consistent performance and flexibility. Different optimization strategies such as particle swarm optimization, differential evolution and Machine Learning algorithms have been utilized in the past to design dipole antennae. This helps in creating a complete device profile and increases its efficacy. Due to the complexity of modern antennas in terms of topology and performance requirements, standard antenna design approaches are tedious and cannot be guaranteed to produce effective results. Out of the strategies that are widely being utilized, Machine Learning (ML) algorithms evolved rapidly due to their capabilities in extrapolating the dimensional and material profiles of the device. Antenna design optimization still faces several difficulties, even though machine learning-based design optimization complements traditional antenna design methodologies. The effectiveness and optimization capabilities of available ML approaches to address a wide range of antenna design problems, considering the increasingly strict specifications of current antennas, are the fundamental difficulties in antenna design optimization which need to be focused on. In our current work, the capability of ML algorithms in elucidating minor trends in device profiles is tested. A bootstrap aggregation model is proposed, concatenating Linear Regression, Support Vector Regression and Decision Tree Regression algorithms. The concatenated model was used to optimize the parameters of reflection coefficient, directivity, efficiency and operating frequency, depending on the feed length, dipole radius and dipole length of the antenna.