Intelligent Design of Antenna for CubeSat Application Using Machine Learning Algorithm

This paper presents an intelligent design approach of a corner truncated microstrip patch antenna (CTMPA) operating at 32 GHz using different well-known machine learning (ML) algorithms. Our goals are to achieve 1) a gain of >6 dBi, 2) an axial ratio (AR) of <3 dB, and 3) a return loss of < -10 across a 10% bandwidth. To do so, we start with a data-set of 715 full-wave simulated samples with 4 different antenna parameters (
$vi{z}$
. features) along with the corresponding computed
$S_{11}$
, gain, and AR. Various ML regression models were analyzed to compare the training data with new predicted values using mean square error (MSE), root mean square error (RMSE), and
$\mathrm{R}^{2}$
score. Then, the best model that meets our goals was chosen. In our work, ten different regression models were analyzed. Among them, the K-nearest neighbor (KNN) regression produced the lowest error. Using the outcome of this model, we predicted the design parameters that achieve our desired goals. The predicted design was verified using full-wave simulation, showing a great agreement.