Model-Driven Channel Estimation for MIMO Monostatic Backscatter System With Deep Unfolding

Abstract

Monostatic backscatter has garnered significant interest due to its distinct benefits in low-cost passive sensing. Observing and sensing with backscatter necessitates determining the phase and amplitude of the backscatter channel to identify the state of the target of interest. In the detection of multiple targets, colliding signals can distort the backscatter channel, complicating channel state recovery. It becomes even more challenging when multiple backscattering devices (BDs) are used. This paper proposes a novel channel estimation scheme to tackle the challenge, which is applied to a monostatic backscatter communication system with multiple reader antennas (RAs) and backscatter devices. Specifically, we propose a backscatter communication model and subsequently develop a de-interfering channel estimation framework that considers the ambient interference in the channel, named model-driven unfolded channel estimation (MUCE). To validate the effectiveness and advantages of the MUCE method, it is compared with the least square (LS) algorithm and convolutional neural network (CNN). The results prove that MUCE requires lower computational costs for the same channel estimation performance and achieves an optimal balance between estimation performance and computational expense.