Single-Antenna Non-Line-of-Sight Matrix Imaging via Reconfigurable Intelligent Surfaces

Abstract

Modern imaging and sensing in complex environments, ranging from biomedical diagnostics to wireless communication, relies on accurately measuring and then controlling the wave propagation. Conventional approaches demand large arrays of antennas or transducers to reconstruct the full reflection or transmission matrix, enabling advanced protocols such as selective focusing or adaptive wave control. Yet, these arrays are expensive, bulky, and difficult to implement at microwave frequencies. Here, we show that a single transmitting-receiving antenna, when combined with a reconfigurable intelligent surface (RIS), can fully reconstruct the reflection matrix from far-field measurements, effectively transforming the RIS into a programmable synthetic antenna array. This approach allows high-fidelity imaging of complex scenes, selective focusing through clutter, and real-time tracking of moving targets. Our results establish RIS as a versatile, low-cost platform for matrix-based imaging, with broad implications for adaptive wave control, real-time sensing, and imaging in environments previously considered inaccessible.

Figures (8)

• Figure 1: Principle of the method. Schematic of the single-antenna system coupled to a RIS for retrieving the reflection matrix in a non-line-of-sight configuration. The antenna’s emitted signal is scattered by the RIS to illuminate the scene, and the back-scattered field—reflected again by the metasurface—is recorded for multiple RIS configurations. From these measurements, the reflection matrix in the RIS-element basis is reconstructed, enabling the computation of the scene image.
• Figure 2: Experimental setup. (A) Photograph of the experiment for the matrix imaging of a single metal target with one $8\times 8$ RIS and a horn antenna. (B) Schematic illustration of matrix imaging with a single antenna and a RIS. We consider a microwave horn antenna illuminating a RIS (B1), the wave then bounces off the RIS toward the scattering scene to image (B2). The wave then travels back to the RIS (B3) before being detected with the same antenna (B4). For the calibration measurement, we remove the scattering medium so that the wave follows only (B1) and (B4).
• Figure 3: Calibration. (A and B) Normalized amplitude and phase respectively of $\mathbf{G}$. Despite being in a room, the wave propagates roughly as in free space. (C) Matrix of interaction between the elements of the RIS $\mathbf{G}_{dd}$. The color represents the phase and the amplitude is coded in the transparency. (D) Amplitude of the coupling between elements as a function of their distance. The red dashed line represents the minimum distance $\rho$ between two elements; the black dashed line corresponds to $1/\rho$ which should be followed for pure radiative coupling. (E) Comparison of the estimated field with the experimental values; the mean absolute error is $5.4\%$ with the initial values and $0.2\%$ after optimization. (F) Image obtained by back-propagating the vector $\mathbf{G}$, which corresponds to the location of the antenna.
• Figure 4: Imaging two targets. (A) Comparison of the estimated field with the experimental values; the mean absolute error is $1.2\%$. (B and C) Respectively, amplitude and phase of the estimated reflection matrix $\mathbf{R}$. (D) Plot of the singular values $\sigma$ of $\mathbf{R}$. (E and F) Phase of the first two singular vectors $\mathbf{U}_1$ and $\mathbf{U}_2$ respectively, associated with the largest singular values. (G) Confocal image obtained from $\mathbf{R}$. (H and I) Images obtained from the first two singular vectors $\mathbf{U}_1$ and $\mathbf{U}_2$ respectively.
• Figure 5: Imaging complex targets. (A) Comparison of the estimated field with the experimental values for the most complex target in the shape of the letter "R". The mean absolute errors for each of the letters of "IETR" are $1.4\%$, $2.1\%$, $1.3\%$ and $3.3\%$ respectively. (B and C) Respectively, amplitude and phase of the estimated reflection matrix $\mathbf{R}$ for the target "R". (D) Singular values $\sigma$ of $\mathbf{R}$ for each target. (E) Images obtained from $\mathbf{R}$ for each target: "I" (E1), "E" (E2), "T" (E3) and "R" (E4).