Shared-Aperture Antenna Decoupling Optimization Method Based on Deep Learning Assistance

This paper aims to address the signal coupling problem of a shared-aperture dual-band dual-circularly polarized microstrip antenna. A decoupling optimization method that combined a convolutional neural network (CNN) with binary particle swarm optimization (BPSO) was proposed. The method introduced pixelated decoupling branches near the antenna feeds, constructed a surrogate model to capture the nonlinear mapping between the branch topology and the electromagnetic performance using a CNN, and adopted BPSO to perform global optimization on the binary pixel matrix, thereby alleviating the time-consuming optimization in a complex, high-dimensional parameter space. Simulation results showed that the optimized S21 was reduced by an average of 20 dB over 1.607–1.620 GHz and by an average of 25 dB over 2.487–2.502 GHz, effectively improving the port isolation. These findings demonstrate that the proposed intelligent optimization strategy is effective and practically applicable for solving antenna decoupling problems.