Deep Denoising Neural Network Assisted Compressive Channel Estimation for mmWave Intelligent Reflecting Surfaces

Abstract

Integrating large intelligent reflecting surfaces (IRS) into millimeter-wave (mmWave) massive multi-input-multi-ouput (MIMO) has been a promising approach for improved coverage and throughput. Most existing work assumes the ideal channel estimation, which can be challenging due to the high-dimensional cascaded MIMO channels and passive reflecting elements. Therefore, this paper proposes a deep denoising neural network assisted compressive channel estimation for mmWave IRS systems to reduce the training overhead. Specifically, we first introduce a hybrid passive/active IRS architecture, where very few receive chains are employed to estimate the uplink user-to-IRS channels. At the channel training stage, only a small proportion of elements will be successively activated to sound the partial channels. Moreover, the complete channel matrix can be reconstructed from the limited measurements based on compressive sensing, whereby the common sparsity of angular domain mmWave MIMO channels among different subcarriers is leveraged for improved accuracy. Besides, a complex-valued denoising convolution neural network (CV-DnCNN) is further proposed for enhanced performance. Simulation results demonstrate the superiority of the proposed solution over state-of-the-art solutions.

Semi-Blind Cascaded Channel Estimation for Reconfigurable Intelligent Surface Aided Massive MIMO

“…1) We formulate the semi-blind cascaded channel estimation problem in the RIS-aided MIMO system as a trilinear estimation problem, where the received signal consists of the product of the RIS-BS channel matrix, the user-RIS channel matrix, and the transmit data matrix. Additionally, unlike the existing training-based works [14]- [22] that estimate the channel coefficients and the transmitted signal separately, we consider a joint estimation of the both channels and the payload data with a small number of pilots . 2) We develop a computationally efficient iterative algorithm to approximately solve the trilinear estimation problem based on the Bayesian minimum mean square error (MMSE) criterion.…”

“…Besides these, machine learning based RIS channel estimation methods were also introduced. For instance, in [21] a twin convolutional neural network architecture was designed to estimate the RIS channels and in [22] a deep denoising neural network was used to assist the estimation of the compressive RIS channels. In addition, [28] considered the channel estimation of wideband RIS-aided orthogonal frequency division multiplexing system.…”

“…Consequently, the transmitter-RIS and RIS-recevier channels cannot be estimated via traditional MIMO channel estimation methods. Currently, the design of the CSI acquisition method in RIS-aided MIMO systems is still in its infancy and the stateof-art approaches (see e.g., [14]- [22]) are pure training-based schemes that require prohibitively long training pilots. This motivates us to explore new channel estimation techniques that reliably estimate the RIS channels with a reduced number of pilot overhead.…”

Interplay between RIS and AI in Wireless Communications: Fundamentals, Architectures, Applications, and Open Research Problems

“…To solve this problem, deep denoising neural networks can be effective. In [66], a hybrid passive/active RIS architecture is first proposed, in which a small part of RIS elements are activated and a few receive chains are deployed to estimate the partial channels. A conventional CS algorithm, i.e., orthogonal match pursuit (OMP) is applied to reconstruct the complete channel matrix whereby the angle domain is sparse.…”

“…After preliminary estimation, a complex-valued denoising convolution neural network (CV-DnCNN) is used to further enhance the estimation accuracy. Similar to [66], the overall channel estimation problem is also divided into two tasks in [67]. The first task is to activate a small number of RIS elements for the angle parameter estimation, and the second task is to utilize a DL framework for further estimation accuracy improvement.…”