Image-driven reflection pattern selection for standalone IRS control

Abstract

Intelligent reflecting surfaces (IRSs), which can dynamically control radio propagation, have attracted attention as devices for achieving non-line-of-sight communication under high-frequency bands. Channel estimation is required to control the IRSs, and the overhead of the estimation reduces the transmission efficiency. To remove channel estimation, control methods to predict reflection pattern candidates from image sensing are proposed. However, searching for the appropriate pattern among these candidates is necessary. This paper proposes an IRS control method that recognize users from camera images and selecting reflection patterns, thereby obviating channel estimation and reflection pattern search. Numerical experiments demonstrate that the proposed method realize accurate IRS control by only camera images.