Enhancing 5G Sub-3.5 GHz Networks through Optimization of Microstrip Antenna Arrays with Sequential Rotation-Based MIMO using Machine Learning Techniques

Abstract

The research concerns the design and optimization of microstrip antenna arrays with sequential rotation-based MIMO configurations for 5G sub-3.5 GHz networks. In total, the use of five optimization approaches was accessed, namely, Bayesian optimization, sparse learning, genetic algorithms, particle swarm optimization, and convolutional neural networks, to evaluate their efficiency in improving antenna properties. In general, Bayesian optimization outperformed the other methodologies most consistently, with the value of the average antenna gain ranging from 0.84 to 0.88, SNR between 14.8 and 15.3, and radiation efficiency ranging between 0.90 and 0.93. The use of sparse learning also ensured a high level of performance, with the average gain ranging between 0.80 and 0.85, and the values of SNR between 15.2 and 15.7. Thus, this approach showed the highest resistance to noise. Genetic algorithms, particle swarm optimization , and CNN are also quite close to each other in the efficiency of properties, with the average antenna gain value ranging between 0.85 and 0.89 and the value of SNR between 14.7 and 15.1. In overall, the results of the research show that there is a variety of available approaches to the optimization of antenna design, and each of them has a number of advantages that can be of use for the improvement of the properties and makes it possible to derive a solution that outweighs others in terms of several properties and the efficiency of its application.