A Wideband Circularly Polarized Antenna with a Non-Uniform Metasurface Designed via Multi-Objective Bayesian Optimization

This is the authors' abstract. We don't add key points for this paper.

A non-uniform metasurface circularly polarized patch antenna is designed using a genuine multi-objective Bayesian optimization (MOBO) method. Instead of aggregating the performance metrics into a single weighted sum, a genuine MOBO approach is adopted to explore the performance trade-offs, where three important antenna performance metrics, input impedance bandwidth, axial ratio (AR) bandwidth, and gain at boresight, are considered as the optimizing targets. Starting from a reference antenna with a uniform metasurface, the non-uniform metasurface antenna is designed by tuning ten physical design parameters automatically. The optimized antenna achieves an operating bandwidth of 25.6% (3.75 GHz–4.85 GHz) with |S11| < -10dB, axial ratio less than 3 dB, and a maximum realized gain of 7.15 dBic. Compared with the reference antenna, the optimized design realizes a 31% enhancement in operating bandwidth while keeping the identical size of 1.12λ 0 × 0.86λ 0 × 0.069λ 0 (λ 0 is the free space wavelength at center frequency of 4.3 GHz). Good agreement between measurement and simulation results validates the optimization methodology.