Machine Learning for Antenna Design, Prediction, and Optimization: A Review

Abstract

Antenna design optimization has attracted significant research interest in recent years, largely because traditional antenna design approaches are often time-consuming and do not always guarantee optimal results. The increasing complexity of modern antennas, in terms of geometry, topology, and strict performance requirements, makes conventional trial-and-error methods less effective. As a result, optimization techniques have become an important complement to classical antenna design methods. However, antenna design optimization still faces several challenges, particularly in achieving high efficiency and strong optimization capability when dealing with complex and highly constrained design problems. In antenna engineering, optimization can involve single-objective techniques, where one performance parameter such as gain, bandwidth, or efficiency is optimized, or multi-objective techniques, where several performance metrics such as gain, bandwidth, isolation, and radiation efficiency are optimized simultaneously. While traditional optimization algorithms have been widely used for these tasks, their computational cost and limited adaptability can restrict their effectiveness for complex antenna structures. Recent advances in machine learning (ML) have introduced new opportunities for improving antenna design optimization. ML-based methods can significantly reduce computational time, improve prediction accuracy, and efficiently explore large design spaces. This paper reviews recent developments in antenna design optimization, with particular emphasis on approaches that integrate machine learning with both single-objective and multi-objective optimization techniques. These emerging methods show strong potential for addressing the growing demands of modern antenna systems and are expected to play an important role in the future development of antennas for a wide range of wireless communication applications.