A review of the most recent advances in deep learning (DL) as applied to electromagnetics (EM), antennas, and propagation is provided. It is aimed at giving the interested readers and pratictioners in EM and related applicative fields some useful insights on the effectiveness and the potentialities of deep neural networks (DNNs) as computational tools with an unprecedented computational efficiency. The range of considered applications includes forward/inverse scattering, direction-of-arrival (DoA) estimation, radar and remote sensing (RS), and multi-input/multi-output (MIMO) systems. Appealing DNN-based solutions concerned with localization, human behavior monitoring (HBM), and EM compatibility (EMC) are reported, as well. Some final remarks are drawn along with indications on future trends according to the authors' viewpoint.

Combining a reflectarray and a transmitarray together, a novel array antenna with bidirectional high-gain beams is proposed in this paper. A novel single-layer metalonly defected square-ring slot element, which has a near-zero thickness, is introduced as the phasing element. Full 360° phase shift range is achieved by utilizing the cross-polarized field of the phasing element. As a bidirectional high-gain antenna (bi-HGA), a transmit-reflect-array using the proposed elements is then designed, fabricated, and measured. It is only composed of a thin metallic sheet with 420 mm side length, spatially fed by a corrugated horn. Well-defined pencil beams are formed in broadside directions on both sides. The measured gains of the transmitted and reflected beams at 10 GHz are 25.5 dBi with 15% 1 dB gain bandwidth and 25 dBi with 14% 1 dB gain bandwidth, respectively. The proposed bi-HGA exhibits distinctive advantages of being ultrathin (0.0067λ ₀ ), lightweight, and low cost without using dielectric substrates, and is a promising candidate for bidirectional wireless communication applications.

Programmable metasurfaces have recently been proposed to dynamically manipulate electromagnetic (EM) waves in both temporal and spatial dimensions. With active components integrated into unit cells of the metasurface, states of the unit cells can be adjusted by digital codes. The metasurface can then construct complex spatial and temporal electromagnetic beams. Given the main parameters of the beam, the optimal codes can be computed by nonlinear optimization algorithms, such as genetic algorithm, particle swarm optimization, etc. The high computational complexity of these algorithms makes it very challenging to compute the codes in real time. In this study, we applied deep learning techniques to compute the codes. A deep convolutional neural network is designed and trained to compute the required element codes in milliseconds, given the requirement of the waveform. The average accuracy of the prediction reaches more than 94 percent. This scheme is validated on a 1-bit programmable metasurface and both experimental and numerical results agree with each other well. This study shows that machines may ”learn” the physics of modulating electromagnetic waves with the help of the good generalization ability in deep convolutional neural networks. The proposed scheme may provide us with a possible solution for real-time complex beam forming in antenna arrays, such as the programmable metasurface.

A novel 1-bit 256-element reconfigurable transmitarray antenna (RTA) at the Ku-band is presented. The element consists of two orthogonal H-shaped slots as receiving and transmitting structures, and the power transmission is realized by a coupling microstrip line in between. Two p-i-n diodes integrated on the coupling line can be electronically controlled to reverse the excitation directions, thus generating two states with 180° phase difference and low transmission loss. A subwavelength element spacing of λ ₀ /3 is utilized to stabilize the element performance under oblique incidence. A transmitarray prototype with 16 × 16 elements and aperture size of 5.3λ ₀ × 5.3λ ₀ at 12.5 GHz is fabricated and measured for the experimental verification. The measured results show that the maximum gain is 17.0 dBi with an aperture efficiency of 14.0%, and 2-D scanning beams within ±50° angular range are obtained. The 3-dB gain bandwidth of the broadside beam is 9.6%.

In this work, we study the application of the supervised descent method (SDM) for two-dimensional microwave imaging. SDM contains offline training and online prediction. In the offline stage, a training data set is generated according to prior information. Then the average descent directions between a fixed initial model and the training models can be learned by iterative schemes. In the online stage, model reconstruction is achieved through iterations based on learned descent directions. This scheme offers a new perspective to incorporate prior information into inversion, and reduce the computational complexity in the online inversion. Synthetic examples validate the accuracy and efficiency of this method.