Coding Programmable Metasurfaces Based on Deep Learning Techniques.

Tao Shan,Xiaotian Pan,Maokun Li,Shenheng Xu,Fan Yang
DOI: https://doi.org/10.1109/jetcas.2020.2972764
IF: 5.877
2020-01-01
IEEE Journal on Emerging and Selected Topics in Circuits and Systems
Abstract:In this paper, we propose a coding scheme based on deep learning techniques for complex beam forming with programmable metasurfaces. A fully convolutional neural network is carefully designed that can "learn" the physics of beam forming from computed data, and make online prediction of the coding matrices. Both numerical and experimental results show that the network can compute coding matrices fulfilling the input requirement in less than one millisecond. This scheme may open a door for real-time complex beam steering with programmable metasurfaces and digital phased arrays.
What problem does this paper attempt to address?