Hybrid Deep Learning Framework For Intelligent Antenna Design Optimization Using Cnn–lstm With Physics-informed Learning

Abstract

Modern Wireless Communication Systems Demand Antennas With High Gain, Wide Bandwidth, And Low Return Loss While Maintaining Compact Size And Reduced Development Time. Conventional Electromagnetic Simulation–based Antenna Design Techniques, Although Accurate, Suffer From High Computational Cost And Prolonged Iterative Optimization Cycles. This Paper Proposes An Intelligent Hybrid Deep Learning Framework That Integrates Convolutional Neural Networks (CNNs) And Long Short-Term Memory (LSTM) Networks Enhanced With An Attention Mechanism And Physics-informed Loss Function For Efficient Antenna Performance Prediction. CNN Layers Extract Spatial Features From Antenna Geometries, While LSTM Networks Model Frequency-dependent Electromagnetic Behavior. The Attention Mechanism Prioritizes Influential Design Parameters, And Physics-informed Constraints Ensure Electromagnetic Validity.

Keywords