ANN-based design of a versatile millimetrewave slotted patch multi-antenna configuration for 5G scenarios

This study addresses the modelling of a dual band (28 and 38 GHz), circularly polarised slotted-patch-antenna for highly demanded millimetre wave multi-input multi-output (MIMO)-systems in fifth generation (5G) networks. A computer-aideddesign model is derived by means of an artificial neural network (ANN) which allows obtaining the physical dimensions of a single-fed antenna, satisfying both near-and far-field goals, without resorting to time-consuming electromagnetic simulation. This mathematical model can be implemented in any CAD-tool, as demonstrated within the framework of advanced design system. This allows, for the first time, to carry out optimisations of strategic importance for future 5G non-linear-radiatingsystems, especially operating at millimetre wave, directly addressing their far-field behaviour. The model performance is validated by some examples and measurement results. A further important advantage of this approach is that the trained ANNmodel can be further adopted to fast, but accurately, investigate the complex relationships between antenna layout and its nearfield and far-field performance, such as the resonance conditions and the polarisation behaviour. Indeed arbitrary orthogonalpolarisations (LHCP/RHCP) have been achieved by the aid of the ANN-model of the same topology. This result can be adopted to implement a combination of two independent radiation patterns for the antenna pair: this feature is attractive for MIMO applications. This is confirmed by measurements showing antenna-coupling reduction with the MIMO-array exploiting polarisation diversity.