A Simple Machine Learning‐Assisted Method for Optimal Antenna Design

ABSTRACT This paper mainly designs the specific optimization process of antenna parameters or topology structure based on the online updated multi‐layer perceptron (MLP) agent model, and introduces principal component analysis (PCA) technology to accelerate the overall optimization process. In addition, to address the problem of mixed optimization of machine learning assisted antenna parameters and topology structure, this paper proposes a new encoding approach, which enables two forms of variables to jointly participate in antenna optimization. The specific approach is to treat the topology structure and antenna parameters as a state and encode them in binary, thereby unifying them into a set of binary variables, which facilitates the introduction of subsequent machine learning and optimization algorithms. By optimizing an ultra‐wideband G‐patch monopole antenna and a microstrip pixel antenna, it has been demonstrated that the proposed method has better optimization performance compared to traditional optimization methods and can be applied to antenna parameters or topology structures.