A State-of-the-Art Survey on Advanced Electromagnetic Design: A Machine-Learning Perspective

Abstract

Research on electromagnetic (EM) components is essential to enabling the design and optimization of such devices as antennas and filters, leading to improved functionality, reduced costs, and enhanced overall performance. This paper presents an overview of recent developments in optimization and design automation techniques for EM-component design and modeling. Limitations of conventional optimization methods are discussed, while the need for novel machine learning techniques capable of handling multiple objectives and large design spaces is highlighted. In this study, existing methods in the literature are reviewed from four viewpoints: structural view, algorithm view, component view, and application view. Different schemes in distinct design stages or applications are examined with advantages and drawbacks laid out for easier comprehension. Finally, to broaden the scope of optimization in the field of EM design and modeling, some prospective trends are pointed out to shed light on emerging research hotspots.

Keywords

• Optimization
• design automation
• deep neural networks
• inverse modeling
• microwave computer-aided design (CAD)