Machine Learning Based Optimization of Multiband LPDA Antenna for Wireless Applications

This paper proposes a new method to predict output parameters from a Log-Periodic Dipole Array (LPDA) antenna (1.2GHz to 11GHz) using different types of machine learning algorithms such as XG-Boost, Random Forest, Decision Tree, and Light-GBM. The identified algorithms are used to predict the antenna response for various communication purposes such as GPS, PCS, IMT-2000, WLAN and Bluetooth. This study has also observed the capabilities of machine learning models to quantify peak gain, return loss (S11) and efficiency of antenna at various operating conditions. Out of all these, Light-GBM yields the best prediction accuracy of 92.46% for peak gain, whereas the S11 parameter is 86.73% with the model of XG-Boost. This ability to combine the individual strengths of different machine learning techniques results in not only a better predictor but also valuable insight into the design optimization for LPDA antennas. The present research could be useful for designing futuristic antenna systems for modern wireless communication systems.