Reflected Beam Metasurface Controlled by Reinforcement Learning

Metasurface is a new type of electromagnetic material based on the generalized Snell theorem . It can achieve different reflection and incidence effects from natural materials, which has attracted extensive attention in the field of electromagnetics. In this paper, a one-dimensional metasurface is designed and simulated which can change the reflected electromagnetic wave beam. The additional phase difference on the metasurface unit is controlled by 1 bit coding to control the direction of the reflected wave beam. Combined with the Deep Q Network algorithm in the reinforcement learning tool, the coding corresponding to the beam meeting the target direction is found.