Inverse Modeling of Quad-Band Notched UWB MIMO Antennas using Gaussian Process Regression

In this paper, a machine learning (ML) model based on Gaussian process regression (GPR) is presented for inverse modeling arrow head-shaped MIMO antenna. The antenna satisfies UWB bandwidth criteria from 3-10.7 GHz and realizes quad band notch characteristics for WiMAX, WLAN and C-band satellite communication systems (downlink and uplink). Datasets required to train the ML model are obtained by varying the dimensions of the MIMO antenna through a finite element method solver. The proposed GPR model is used to estimate the geometrical parameters of the MIMO antenna using cut-off frequencies and four notch frequencies. An artificial neural network (ANN) model based on multilayer perceptron (MLP) is also proposed for comparative analysis. The results obtained using GPR and MLP are compared and it is observed that GPR has outperformed the MLP model in terms of various statistical measures.