Ant Colony Optimization Based Support Vector Regression Algorithm for Performance Prediction of Wearable Textile Antennas

Body-centric wireless communication has been made possible by textile patch antennas which have lately gained significance in wearable applications. They are used in a wide range of industries, including healthcare and defence. However, due to varying properties of substrate materials and complexity in the geometric features of patch antennas, it is often challenging to optimize and predict their performance in terms of gain, directivity and return loss. In this paper, an ant colony optimization-based support vector machine regression algorithm (ACO-SVM) is proposed for modelling the relationship between different antenna parameters and output performance (in terms of gain, directivity and return loss). An ablation study is performed to explore the role of operating frequency in determining the output signal strength of antennas. Experiments show that the ACO-SVM models this complex relation most accurately with the lowest loss values: Mean Squared Error (MSE) = 5.33, Mean Absolute Error (MAE) = 0.22, Root Mean Squared Error (RMSE) = 1.02, R
²
Score = 0.13) and inference time of 2ms. To validate the model’s performance, a textile patch antenna is designed using HFSS with a range of fabric-based substrate materials (cotton, poly-amide, and polyester) and its performance is analysed in simulation. It is shown that the simulated results and the algorithmic results obtained with ACO-SVM agree well with one another and that the proposed ACO-SVM approach can predict optimal antenna design parameters given the required performance in terms of gain and return loss.