Coding 1‐Bit Programmable Metasurface Based on Bayesian Learning With a Hierarchical Truncated Gaussian Mixture Prior Model

ABSTRACT Emerging programmable metasurface technologies enable flexible manipulation of electromagnetic waves. However, the instantaneous design of corresponding code sequences remains a challenging problem. In this paper, a fast and reliable coding strategy for 1‐bit programmable metasurfaces is proposed. Under the assumption that the reflection amplitudes for both ‘0’ and ‘1’ coding states are equal to unity, the original coding problem is first formulated as a linear inverse problem. Since the phase difference between ‘0’ and ‘1’ coding states is approximately , such a prior information is then exploited for addressing the linear inverse problem. Specifically, a hierarchical truncated Gaussian mixture (TGM) prior model is adopted to account for the feature of the unknowns. Moreover, the variational expectation maximisation (VEM) along with the generalised approximate message passing (GAMP) is used for Bayesian inference, significantly reducing the computational complexity. Several numerical examples are presented to illustrate the potential of the proposed coding scheme.