Deep-learning-assisted reconfigurable metasurface antenna for real-time holographic beam steering

Abstract We propose a metasurface antenna capable of real-time holographic beam steering. An array reconfigurable dipoles can generate on-demand far-field patterns radiation through the specific encoding meta-atomic states i.e., configuration each dipole. Suitable for generation desired be identified using iteration, but this is very slow and needs to done pattern. Here, we present deep-learning-based method control with point dipole elements that vary in their state polarizability. Instead adopt deep learning algorithm combines an autoencoder electromagnetic scattering equation determine required target pattern real-time. The from Born approximation used as decoder training neural network, analytic Green’s function calculation check validity approximation. Our learning-based requires computing time within 200 μs states, thus enabling operation antenna.