Joint Sparse Estimation with Cardinality Constraint via Mixed-Integer Semidefinite Programming

The multiple measurement vectors (MMV) problem refers to the joint estimation of a row-sparse signal matrix from multiple realizations of mixtures with a known dictionary. As a generalization of the standard sparse representation problem for a single measurement, this problem is fundamental in various applications in signal processing, e.g., spectral analysis and direction-of-arrival (DOA) estimation. In this paper, we consider the maximum a posteriori (MAP) estimation for the MMV problem, which is classically formulated as a regularized least-squares (LS) problem with an $\ell_{2,0}$-norm constraint, and derive an equivalent mixed-integer semidefinite program (MISDP) reformulation. The proposed MISDP reformulation can be exactly solved by a generic MISDP solver, which, however, becomes computationally demanding for problems of extremely large dimensions. To further reduce the computation time in such scenarios, a relaxation-based approach can be employed to obtain an approximate solution of the MISDP reformulation, at the expense of a reduced estimation performance. Numerical simulations in the context of DOA estimation demonstrate the improved error performance of our proposed method in comparison to several popular DOA estimation methods. In particular, compared to the deterministic maximum likelihood (DML) estimator, which is often used as a benchmark, the proposed method applied with a state-of-the-art MISDP solver exhibits a superior estimation performance at a significantly reduced running time. Moreover, unlike other nonconvex approaches for the MMV problem, including the greedy methods and the sparse Bayesian learning, the proposed MISDP-based method offers a guarantee of finding a global optimum.

SALSA: A Sequential Alternating Least Squares Approximation Method For MIMO Channel Estimation

In this paper, we consider the channel estimation problem in sub-6 GHz uplink wideband MIMO-OFDM communication systems, where a user equipment with a fully-digital beamforming structure is communicating with a base station having a hybrid analog-digital beamforming structure. A novel channel estimation method called Sequential Alternating Least Squares Approximation (SALSA) is proposed by exploiting a hidden tensor structure in the uplink measurement matrix. Specifically, by showing that any MIMO channel matrix can be approximately decomposed into a summation of R factor matrices having a Kronecker structure, the uplink measurement matrix can be reshaped into a 3-way tensor admitting a Tucker decomposition. Exploiting the tensor structure, the MIMO channel matrix is estimated sequentially using an alternating least squares method. Detailed simulation results are provided showing the effectiveness of the proposed SALSA method as compared to the classical least squares method.

The Perfect Match: RIS-enabled MIMO Channel Estimation Using Tensor Decomposition

The deployment of reconfigurable intelligent surfaces (RISs) in a communication system provides control over the propagation environment, which facilitates the augmentation of a multitude of communication objectives. As these performance gains are highly dependent on the applied phase shifts at the RIS, accurate channel state information at the transceivers is imperative. However, not only do RISs traditionally lack signal processing capabilities, but their end-to-end channels also consist of multiple components. Hence, conventional channel estimation (CE) algorithms become incompatible with RIS-aided communication systems as they fail to provide the necessary information about the channel components, which are essential for a beneficial RIS configuration. To enable the full potential of RISs, we propose to use tensor-decomposition-based CE, which facilitates smart configuration of the RIS by providing the required channel components. We use canonical polyadic (CP) decomposition, that exploits a structured time domain pilot sequence. Compared to other state-of-the-art decomposition methods, the proposed Semi-Algebraic CP decomposition via Simultaneous Matrix Diagonalization (SECSI) algorithm is more time efficient as it does not require an iterative process. The benefits of SECSI for RIS-aided networks are validated with numerical results, which show the improved individual and end-to-end CE accuracy of SECSI.

Combining AI/ML and PHY Layer Rule Based Inference -- Some First Results

In 3GPP New Radio (NR) Release 18 we see the first study item starting in May 2022, which will evaluate the potential of AI/ML methods for Radio Access Network (RAN) 1, i.e., for mobile radio PHY and MAC layer applications. We use the profiling method for accurate iterative estimation of multipath component parameters for PHY layer reference, as it promises a large channel prediction horizon. We investigate options to partly or fully replace some functionalities of this rule based PHY layer method by AI/ML inferences, with the goal to achieve either a higher performance, lower latency, or, reduced processing complexity. We provide first results for noise reduction, then a combined scheme for model order selection, compare options to infer multipath component start parameters, and, provide an outlook on a possible channel prediction framework.

Transfer Learning Capabilities of Untrained Neural Networks for MIMO CSI Recreation

Machine learning (ML) applications for wireless communications have gained momentum on the standardization discussions for 5G advanced and beyond. One of the biggest challenges for real world ML deployment is the need for labeled signals and big measurement campaigns. To overcome those problems, we propose the use of untrained neural networks (UNNs) for MIMO channel recreation/estimation and low overhead reporting. The UNNs learn the propagation environment by fitting a few channel measurements and we exploit their learned prior to provide higher channel estimation gains. Moreover, we present a UNN for simultaneous channel recreation for multiple users, or multiple user equipment (UE) positions, in which we have a trade-off between the estimated channel gain and the number of parameters. Our results show that transfer learning techniques are effective in accessing the learned prior on the environment structure as they provide higher channel gain for neighbouring users. Moreover, we indicate how the under-parameterization of UNNs can further enable low-overhead channel state information (CSI) reporting.

Machine Learning for CSI Recreation Based on Prior Knowledge

Knowledge of channel state information (CSI) is fundamental to many functionalities within the mobile wireless communications systems. With the advance of machine learning (ML) and digital maps, i.e., digital twins, we have a big opportunity to learn the propagation environment and design novel methods to derive and report CSI. In this work, we propose to combine untrained neural networks (UNNs) and conditional generative adversarial networks (cGANs) for MIMO channel recreation based on prior knowledge. The UNNs learn the prior-CSI for some locations which are used to build the input to a cGAN. Based on the prior-CSIs, their locations and the location of the desired channel, the cGAN is trained to output the channel expected at the desired location. This combined approach can be used for low overhead CSI reporting as, after training, we only need to report the desired location. Our results show that our method is successful in modelling the wireless channel and robust to location quantization errors in line of sight conditions.

Compressed Sensing Constant Modulus Constrained Projection Matrix Design and High-Resolution DoA Estimation Methods

This paper proposes a compressed sensing-based high-resolution direction-of-arrival estimation method called gradient orthogonal matching pursuit (GOMP). It contains two main steps: a sparse coding approximation step using the well-known OMP method and a sequential iterative refinement step using a newly proposed gradient-descent method. To enhance the recoverability, we further propose an efficient projection matrix design method, which considers the constant modulus constraints imposed by the projection matrix hardware components. Simulation results show the effectiveness of the proposed methods as compared to benchmark algorithms.

Robust Multi-dimensional Model Order Estimation Using LineAr Regression of Global Eigenvalues (LaRGE)

The efficient estimation of an approximate model order is very important for real applications with multi-dimensional data if the observed low-rank data is corrupted by additive noise. In this paper, we present a novel robust method for model order estimation of noise-corrupted multi-dimensional low-rank data based on the LineAr Regression of Global Eigenvalues (LaRGE). The LaRGE method uses the multi-linear singular values obtained from the HOSVD of the measurement tensor to construct global eigenvalues. In contrast to the Modified Exponential Test (EFT) that also exploits the approximate exponential profile of the noise eigenvalues, LaRGE does not require the calculation of the probability of false alarm. Moreover, LaRGE achieves a significantly improved performance in comparison with popular state-of-the-art methods. It is well suited for the analysis of biomedical data. The excellent performance of the LaRGE method is illustrated via simulations and results obtained from EEG recordings.

Double-RIS Versus Single-RIS Aided Systems: Tensor-Based MIMO Channel Estimation and Design Perspectives

Reconfigurable intelligent surfaces (RISs) have been proposed recently as new technology to tune the wireless propagation channels in real-time. However, most of the current works assume single-RIS (S-RIS)-aided systems, which can be limited in some application scenarios where a transmitter might need a multi-RIS-aided channel to communicate with a receiver. In this paper, we consider a double-RIS (D-RIS)-aided MIMO system and propose an alternating least-squared-based channel estimation method by exploiting the Tucker2 tensor structure of the received signals. Using the proposed method, the cascaded MIMO channel parts can be estimated separately, up to trivial scaling factors. Compared with the S-RIS systems, we show that if the RIS elements of a S-RIS system are distributed carefully between the two RISs in a D-RIS system, the training overhead can be reduced and the estimation accuracy can also be increased. Therefore, D-RIS systems can be seen as an appealing approach to further increase the coverage, capacity, and efficiency of future wireless networks compared to S-RIS systems.

Tensor-Based Channel Estimation and Reflection Design for RIS-Aided Millimeter-Wave MIMO Communication Systems

In this work, we consider both channel estimation and reflection design problems in point-to-point reconfigurable intelligent surface (RIS)-aided millimeter-wave (mmWave) MIMO communication systems. First, we show that by exploiting the low-rank nature of mmWave MIMO channels, the received training signals can be written as a low-rank multi-way tensor admitting a canonical polyadic CP decomposition. Utilizing such a structure, a tensor-based RIS channel estimation method (termed TenRICE) is proposed, wherein the tensor factor matrices are estimated using an alternating least squares method. Using TenRICE, the transmitter-to-RIS and the RIS-to-receiver channels are efficiently and separately estimated, up to a trivial scaling factor. After that, we formulate the beamforming and RIS reflection design as a spectral efficiency maximization problem. Due to its non-convexity, we propose a heuristic non-iterative two-step method, where the RIS reflection vector is obtained in a closed form using a Frobenius-norm maximization (FroMax) strategy. Our numerical results show that TenRICE has a superior performance, compared to benchmark methods, approaching the Cram\'er-Rao lower bound with a low training overhead. Moreover, we show that FroMax achieves a comparable performance to benchmark methods with a lower complexity.