Tensor Completion-Based Channel Estimation for Semi-passive RIS Assisted System

Previous works mainly considered semi-passive Re-configurable intelligent surface (RIS) design with special layout arrangements, ignoring the fact that the local observed values may not reflect the overall channel when RIS elements increase. In this paper, we design a semi-passive RIS structure with a random arrangement, and propose a tensor completion-based channel estimation algorithm to recover the whole channel from the partially observed signals. Specifically, we introduce the tensor singular value decomposition (t-svd) framework to learn the inherent low-rank representation of the observed data: the search for inherent basis representations is carried out on the t-Grassmannian manifold, and the representation of low-rank tensor under this basis has a closed-form solution. As long as the proportion of active components reaches a certain level, the proposed algorithm can work well. Simulations show that the t-svd-based tensor completion algorithm performs better than the CP decomposition-based tensor completion algorithm.