This paper studies the multi-access coded caching (MACC) problem with arbitrary user-cache access topology, which extends existing MACC models that rely on highly structured and combinatorially designed topologies. We consider a MACC system consisting of a single server, $Λ$ cache-nodes, and $K$ user-nodes. The server stores $N$ equal-size files, each cache-node has a storage capacity of $M$ files, and each user-node $k\in[K]$ can access an arbitrary subset of cache-nodes $\mathcal{A}_k\subseteq[Λ]$ and retrieve the cached content stored in cache-nodes $\mathcal{A}_k$. The objective is to design a universal framework for the MACC delivery problem. Decoding conflicts among the requested packets are captured by a conflict graph, and the design of the delivery is reduced to a graph coloring problem, where achieving a lower transmission load corresponds to coloring the graph using fewer colors. Under this formulation, the classical DSatur algorithm achieves a transmission load close to the index-coding (IC) converse bound, thereby providing a practical benchmark. However, its computational complexity becomes prohibitive for large-scale graphs. To overcome this limitation, we develop a learning-driven approach using graph neural networks (GNNs) that efficiently constructs coded multicast transmissions with performance close to the theoretical bounds and generalizes across different user-cache access topologies and numbers of users. In addition, we extend the IC converse bound to MACC systems with arbitrary access topology and propose a low-complexity greedy approximation that closely matches the IC converse bound. Numerical results demonstrate that the proposed approach achieves performance close to the DSatur algorithm and the IC converse bound, while significantly reducing computational complexity, making it well-suited for large-scale MACC systems.

Integrated sensing and communication (ISAC) systems provide significant enhancements in performance and resource efficiency compared to individual sensing and communication systems, primarily attributed to the collaborative use of wireless resources, radio waveforms, and hardware platforms. This paper focuses on the bistatic ISAC systems with separated multi-receiver and one sensor. Compared to a monostatic ISAC system, the main challenge in the bistatic setting is that the information messages are unknown to the sensor and therefore they are seen as interference, while the channel between the transmitters and the sensor is unknown to the transmitters. In order to mitigate the interference at the sensor while maximizing the communication degree of freedom, we introduce two strategies, namely, blind interference alignment and topological interference management. Although well-known in the context of Gaussian interference channels, these strategies are novel in the context of bistatic ISAC. For the bistatic ISAC models with heterogeneous coherence time or with heterogeneous connectivity, the achieved ISAC tradeoff points in terms of communication and sensing degrees of freedom are characterized. In particular, we show that the new tradeoff outperforms the time-sharing between the sensing-only and the communication-only schemes. Simulation results demonstrate that the proposed schemes significantly improve the channel estimation error for the sensing task, compared to treating interference as noise at the sensor and successive interference cancellation.

In sixth-generation and beyond, space-air-ground integrated networks (SAGINs) extend network connectivity to space, thereby enabling broader service coverage. This paper proposes a topology-aware SAGIN framework to address the integrated sensing, communication, and wireless power transfer (ISCPT) problem, leveraging the distinctive visibility of satellite-terrestrial and satellite-satellite users as well as their constructing in-between channel strengths. By modeling the topology of the SAGIN as a bipartite graph, we formulate the ISCPT problem as a multi-objective joint optimization problem with specified topological structures to reflect connection relationships of satellite-terrestrial and satellite-satellite users. The ISCPT problem is then reformulated and carefully decomposed as several mixed-integer linear programs (MILPs) by leveraging the network topology to individually optimize sensing, communication, and power transfer. To reduce the computational complexity of the proposed method, a greedy algorithm deal with generalized multi-assignment problem (GMAP) is developed. Simulation results demonstrate superior performance in communication and sensing, with a tolerable trade-off in wireless power transfer.

The upper mid-band balances coverage and capacity for the future cellular systems and also embraces extremely large-scale multiple-input multiple-output (XL-MIMO) systems, offering enhanced spectral and energy efficiency. However, these benefits are significantly degraded under mobility due to channel aging, and further exacerbated by the unique near-field (NF) and spatial non-stationarity (SnS) propagation in such systems. To address this challenge, we propose a novel channel prediction approach that incorporates dedicated channel modeling, probabilistic representations, and Bayesian inference algorithms for this emerging scenario. Specifically, we develop tensor-structured channel models in both the spatial-frequency-temporal (SFT) and beam-delay-Doppler (BDD) domains, which capture the NF and SnS propagation effects and leverage temporal correlations among multiple snapshots for channel prediction. In this model, the factor matrices of multi-linear transformations are parameterized by BDD domain grids and SnS factors, where beam domain grids are jointly determined by angles and slopes under spatial-chirp based NF representations. To enable tractable inference, we replace these environment-dependent BDD domain grids with uniformly sampled ones, and introduce perturbation parameters in each domain to mitigate grid mismatch.We further propose a hybrid beam domain strategy that integrates angle-only sampling with slope hyperparameterization to avoid the computational burden of explicit slope sampling. On this basis, we develop tensor-structured bi-layer inference (TS-BLI) algorithm under the expectation-maximization (EM) framework, which reduces the computational complexity by leveraging the inherent separation across different domains. In the E-step, we develop the bi-layer factor graph representation to isolate the bilinear mixing in the spatial domain induced by SnS propagation, thus facilitating bi-layer iterations using approximate inference techniques. In the M-step, we leverage an alternating strategy for hyperparameter learning, with closed-form rules derived by the quadratic approximation of objective functions. Numerical simulations based on a near-practical channel simulator developed upon QuaDRiGa with SnS extensions demonstrate the superior channel prediction performance of the proposed algorithm.

Despite impressive capability in learning over graphstructured data, graph neural networks (GNN) suffer from adversarial topology perturbation in both training and inference phases. While adversarial training has demonstrated remarkable effectiveness in image classification tasks, its suitability for GNN models has been doubted until a recent advance that shifts the focus fromtransductivetoinductivelearning. Still, GNN robustness in the inductive setting is under-explored, and it calls for deeper understanding of GNN adversarial training. To this end, we introduce a concept of graph subspace energy (GSE)—a generalization of graph energy that measures graph stability—of the adjacency matrix, as an indicator of GNN robustness against topology perturbations. To further demonstrate the effectiveness of such concept, we propose an adversarial training method with the perturbed graphs generated by maximizing the GSE regularization term, referred to as AT-GSE. To deal with the local and global topology perturbations raised respectively by LRBCD and PRBCD, we employ randomized SVD (RndSVD) and Nyström low-rank approximation to favor the different aspects of the GSE terms. An extensive set of experiments shows that AT-GSE outperforms consistently the state-of-the-art GNN adversarial training methods over different homophily and heterophily datasets in terms of adversarial accuracy, whilst more surprisingly achieving a superior clean accuracy on non-perturbed graphs.

In moderate- to high-mobility scenarios, channel state information (CSI) varies rapidly and becomes temporally non-stationary, leading to severe performance degradation in the massive multiple-input multiple-output (MIMO) transmissions. To address this issue, we propose a tensor-structured approach to dynamic channel prediction (TS-DCP) for massive MIMO systems with temporal non-stationarity, exploiting both dual-timescale and cross-domain correlations. Specifically, due to inherent spatial consistency, non-stationary channels over long-timescales can be approximated as stationary on short-timescales, decoupling complicated temporal correlations into more tractable dual-timescale ones. To exploit such property, we propose the sliding frame structure composed of multiple pilot orthogonal frequency-division multiplexing (OFDM) symbols, which capture short-timescale correlations within frames by Doppler domain modeling and long-timescale correlations across frames by Markov/autoregressive processes. Building on this, we develop the Tucker-based spatial-frequency-temporal domain channel model, incorporating angle-delay-Doppler (ADD) domain channels and factor matrices parameterized by ADD domain grids. Furthermore, we model cross-domain correlations of ADD domain channels within each frame, induced by clustered scattering, through the Markov random field and tensor-coupled Gaussian distribution that incorporates high-order neighborhood structures. Following these probabilistic models, we formulate the TS-DCP problem as variational free energy (VFE) minimization, and unify different inference rules through the structure design of trial beliefs. This formulation results in the dual-layer VFE optimization process and yields the online TS-DCP algorithm, where the computational complexity is reduced by exploiting tensor-structured operations. Numerical simulations demonstrate the significant superiority of the proposed algorithm over benchmarks in terms of channel prediction performance.

Integrated sensing and communication (ISAC) systems provide significant enhancements in performance and resource efficiency compared to individual sensing and communication systems, primarily attributed to the collaborative use of wireless resources, radio waveforms, and hardware platforms. The performance limits of a system are crucial for guiding its design; however, the performance limits of ISAC systems remain an open question. This paper focuses on the bistatic ISAC systems with dispersed multi-receivers and one sensor. Compared to the monostatic ISAC systems, the main challenge is that that the communication messages are unknown to the sensor and thus become its interference, while the channel information between the transmitters and the sensor is unknown to the transmitters. In order to mitigate the interference at the sensor while maximizing the communication degree of freedom, we introduce the blind interference alignment strategy for various bistatic ISAC settings, including interference channels, MU-MISO channels, and MU-MIMO channels. Under each of such system, the achieved ISAC tradeoff points by the proposed schemes in terms of communication and sensing degrees of freedom are characterized, which outperforms the time-sharing between the two extreme sensing-optimal and communication-optimal points. Simulation results also demonstrate that the proposed schemes significantly improve on the ISAC performance compared to treating interference as noise at the sensor.

Spectrum sharing in Device-to-Device (D2D) communications with power control and link scheduling is a challenging non-convex combinatorial optimization problem. The state-of-the-art model-based iterative algorithms such as FPLinQ produce optimum-achieving solutions, whilst deep learning-based approaches have been proposed recently to approximate FPLinQ with reduced computational complexity. However, due to the highly non-convex nature of the optimization problem, FPLinQ exhibits certain deficiencies in the highly interference-limited networks, as it may be trapped within certain local sub-optimal solutions that may be far from the global optimum. To address these issues, we propose to unfold FPLinQ, with certain parameters inside the iterative procedure adjusted by a graph reinforcement learning (GRL) method, and end up with a novel hybrid model/data-driven approach, termed UFPLinQ. Not only does UFPLinQ inherit the advantages of FPLinQ and GRL with respect to local optimality, explainability, scalability, and generalizability, but it also provides excellent solutions in interference-limited networks where FPLinQ fails. By numerical evaluations, UFPLinQ outperforms existing learning-based power control mechanisms, with substantially reduced training samples and iterations, and more interestingly remedies the potential deficiencies of FPLinQ in highly interference-limited networks.

This paper presents a dual transformer-based robust precoding scheme for massive multiple-input multiple-output systems with low computational complexity. By utilizing the a posteriori channel model, the imperfect channel state information (CSI) is modeled as the statistical CSI that incorporates channel mean and channel variance information with spatial correlation. Based on this, we formulate a robust precoding problem aimed at maximizing the expected sum rate and subsequently transform it into a robust weighted minimum mean square error problem. We prove the permutation equivariance and invariance satisfied by the mapping from the available CSI to the low-dimensional variables in the optimal closed-form solution. To fully exploit such properties, we design a dual transformer block with residual connection and introduce an invariant transformer module to construct a neural network, which is trained to approximate the mapping for precoding computation. Simulation results demonstrate that this method exhibits strong robustness, lower computational complexity, and high scalability in dynamic user/antenna scenarios compared to other approaches.

Adversarial training has emerged as a highly effective way to improve the robustness of deep neural networks (DNNs). It is typically conceptualized as a min-max optimization problem over model weights and adversarial perturbations, where the weights are optimized using gradient descent methods, such as SGD. In this paper, we propose a novel approach by treating model weights as random variables, which paves the way for enhancing adversarial training through Second-Order Statistics Optimization (S$^{2}$2O) over model weights. We challenge and relax a prevalent, yet often unrealistic, assumption in prior PAC-Bayesian frameworks: the statistical independence of weights. From this relaxation, we derive an improved PAC-Bayesian robust generalization bound. Our theoretical developments suggest that optimizing the second-order statistics of weights can substantially tighten this bound. We complement this theoretical insight by conducting an extensive set of experiments that demonstrate that S$^{2}$2O not only enhances the robustness and generalization of neural networks when used in isolation, but also seamlessly augments other state-of-the-art adversarial training techniques.

We address the recovery of block sparse signals by proposing a distributed solution that uses a block-diagonal approximation to the dictionary matrix of the problem. The approximation is found in two stages. First, the Gram matrix of the dictionary matrix is used as a basis for spectral clustering. Afterwards, measurement positions are assigned to the clusters formed from this spectral clustering. The method is then applied to use previous algorithms in the literature of Block Sparse Bayesian Learning in parallel. Moreover, this method also speeds up the algorithm in serial systems. The efficacy of the proposed method is demonstrated in simulations with comparison to the previous Block Sparse Bayesian Learning algorithms.

In this paper, we consider a radio resource management (RRM) problem in the dynamic wireless networks, comprising multiple communication links that share the same spectrum resource. To achieve high network throughput while ensuring fairness across all links, we formulate a resilient power optimization problem with per-user minimum-rate constraints. We obtain the corresponding Lagrangian dual problem and parameterize all variables with neural networks, which can be trained in an unsupervised manner due to the provably acceptable duality gap. We develop a meta-learning approach with graph neural networks (GNNs) as parameterization that exhibits fast adaptation and scalability to varying network configurations. We formulate the objective of meta-learning by amalgamating the Lagrangian functions of different network configurations and utilize a first-order meta-learning algorithm, called Reptile, to obtain the meta-parameters. Numerical results verify that our method can efficiently improve the overall throughput and ensure the minimum rate performance. We further demonstrate that using the meta-parameters as initialization, our method can achieve fast adaptation to new wireless network configurations and reduce the number of required training data samples.

In this paper, we investigate the channel estimation and user localization problems for multi-user integrated sensing and communication (ISAC) systems empowered by the reconfigurable intelligent surface (RIS) technology. In order to perceive environmental information more deeply, we propose a spherical RIS architecture with spherically arranged unit cells. Based on the principle of phase mode excitation, we customize the design of RIS profiles and recover the equivalent channel parameters via subspace estimation tools. By exploring the characteristics of RIS array manifold and free-space propagation, we develop a decoupling framework of three-dimensional channel parameters, which is not supported by conventional planar RIS topologies. Each user can achieve a self-localization by analyzing the signals transmitted from other active users. Simulation results indicate that the spherical RIS can enable joint channel estimation, user localization and data transmission with remarkable performance that approaches the theoretical Cramér-Rao bounds.

In this paper, we investigate the channel estimation and localization problems for integrated sensing and communication (ISAC) systems empowered by the reconfigurable intelligent surface (RIS) technology. We propose a cylindrical RIS architecture that arranges reflecting elements on a curved substrate, where the three-dimensional array manifold can not only offer a 360° coverage but also perceive the environmental information more deeply. The conformal RIS topology can fit the deployment scenarios more flexibly, which, however, incurs a potential issue of shadowing effect, i.e., signal waves from/to certain directions can only be observed by a part of reflectors due to the shielding of the substrate curvature, yielding different visibility regions (VRs) for multiple users on the RIS array manifold. In order to address this problem, we propose a tensorial channel estimation approach, where the cascaded channel is transformed into the beamspace domain and modeled as a canonical polyadic tensor. By leveraging the principle of tensor completion, we can eliminate the RIS training profiles to deconstruct the channel in the element domain. Then, we develop a VR detection strategy based on the sliding windows, retrieving equivalent channel parameters from the effective signal responses. Finally, by exploring the characteristics of the cylindrical RIS architecture, we develop a decoupling framework to uniquely recover the exact channel parameters, based on which each user can locate itself and other interacting ones. Simulation results indicate that the proposed cylindrical RIS can enable the channel estimation, user localization and data transmission simultaneously, exhibiting remarkable performance under the shadowing effect interference.

Device-to-device (D2D) spectrum sharing in wireless communications is a challenging non-convex combinatorial optimization problem, involving entangled link scheduling and power control in a large-scale network. The state-of-the-art methods, either from a model-based or a data-driven perspective, exhibit certain limitations such as the critical need for channel state information (CSI) and/or a large number of (solved) instances (e.g., network layouts) as training samples. To advance this line of research, we propose a novel hybrid model/data-driven spectrum sharing mechanism with graph reinforcement learning for link scheduling (GRLinQ), injecting information theoretical insights into machine learning models, in such a way that link scheduling and power control can be solved in an intelligent manner. Through an extensive set of experiments, GRLinQ demonstrates superior performance to the existing model-based and data-driven link scheduling and/or power control methods, with a relaxed requirement for CSI, a substantially reduced number of unsolved instances as training samples, a possible distributed deployment, reduced online/offline computational complexity, and more remarkably excellent scalability and generalizability over different network scenarios and system configurations.

The advance of topological interference management (TIM) has been one of the driving forces of recent developments in network information theory. However, state-of-the-art coding schemes for TIM are usually handcrafted for specific families of network topologies, relying critically on experts’ domain knowledge and sophisticated treatments. The lack of systematic and automatic generation of solutions inevitably restricts their potential wider applications to wireless communication systems, due to the limited generalizability of coding schemes to wider network configurations. To address such an issue, this work makes the first attempt to advocate revisiting topological interference alignment (IA) from a novel learning-to-code perspective. Specifically, we recast the one-to-one and subspace IA conditions as vector assignment policies and propose a unifying learning-to-code on graphs (LCG) framework by leveraging graph neural networks (GNNs) for capturing topological structures and reinforcement learning (RL) for decision-making of IA beamforming vector assignment. Interestingly, the proposed LCG framework is capable of recovering known one-to-one scalar/vector IA solutions for a significantly wider range of network topologies, and more remarkably of discovering new subspace IA coding schemes for multiple-antenna cases that are challenging to be handcrafted. The extensive experiments demonstrate that the LCG framework is an effective way to automatically produce systematic coding solutions to the TIM instances with arbitrary network topologies, and at the same time, the underlying learning algorithm is efficient with respect to online inference time and possesses excellent generalizability and transferability for practical deployment.

The increasing demands for spectral resources in emerging massive machine-type communication applications necessitate the implementation of massive grant-free transmission in the millimeter-wave (mmWave) band. This paper proposes two efficient receive analog beamforming design algorithms for mmWave massive grant-free transmission under hybrid beamforming architectures, intending to optimize spectral efficiency and access probability, respectively. Specifically, we first express the spectral efficiency of mmWave massive grant-free transmission systems and then derive an analytically tractable approximation using the random matrix theory. Following this, an alternating optimization method is employed to design the receive beamforming matrix efficiently. Additionally, we provide the formulation of access probability for mmWave massive grant-free transmission, whose explicit expression is approximately derived through the Gaussian approximation. Building upon this, we utilize a convex hull relaxation-based optimization method to optimize the beamforming matrix. The effectiveness of our proposed beamforming design algorithms in improving spectral efficiency and access probability is validated through extensive simulation experiments.

With the increasing demand for spectrum efficiency and energy efficiency, reconfigurable intelligent surfaces (RISs) have attracted massive attention due to its low-cost and capability of controlling wireless environment. However, there is still a lack of treatments to deal with the growth of the number of users and RIS elements, which may incur performance degradation or computational complexity explosion. In this paper, we investigate the joint optimization of user scheduling and precoding for distributed RIS-aided communication systems. Firstly, we propose an optimization-based numerical method to obtain suboptimal solutions with the aid of the approximation of ergodic sum rate. Secondly, to reduce the computational complexity caused by the high dimensionality, we propose a data-driven scalable and generalizable multi-agent deep reinforcement learning (MADRL) framework with the aim to maximize the ergodic sum rate approximation through the cooperation of all agents. Further, we propose a novel dynamic working process exploiting the trained MADRL algorithm, which enables distributed RISs to configure their own passive precoding independently. Simulation results show that our algorithm substantially reduces the computational complexity by a time reduction of three orders of magnitude at the cost of 3% performance degradation, compared with the optimization-based method, and achieves 6% performance improvement over the state-of-the-art MADRL algorithms.

The Invariant Risk Minimization (IRM) approach aims to address the security challenge of out-of-distribution robustness (domain generalization) by training a feature representation that remains invariant across multiple environments. However, in noisy environments, noise can distort invariant features, leading to different environment-specific losses. Current IRM-related methods such as IRMv1 and VREx underperform in these settings because they enforce uniform losses across environments. While environmental noise causes environment-specific losses, it does not alter the fundamental correlation between invariant representations and labels. Based on this observation, we propose ICorr (Invariant Correlation), which leverages this correlation to extract invariant representations in noisy settings. Unlike existing approaches, ICorr accommodates different environment-specific inherent losses while maintaining a necessary condition for identifying IRM classifiers. We present a detailed case study demonstrating why previous methods may lose ground while ICorr can succeed. Through a theoretical lens, particularly from a causality perspective, we illustrate that the invariant correlation of representation with label is a necessary condition for the optimal invariant predictor in noisy environments, whereas the optimization motivations for other methods may not be. Furthermore, we empirically demonstrate the effectiveness of ICorr by comparing it with other domain generalization methods on various noisy datasets.

Deep learning-based Wi-Fi sensing has received massive interest thanks to the prevalence of Wi-Fi technology. While deep learning techniques provide promising results in Wi-Fi sensing, there are only very few studies on the vulnerabilities against Wi-Fi ensing. In this paper, we studied evasion attacks against deep learning-based Wi-Fi sensing and the countermeasure and conducted an extensive experimental evaluation using two publicly available datasets, namely SignFi and Widar. Accordingly, we proposed three white-box and two black-box attacks and revealed that even with an undetectable power change, evasion attacks can achieve a remarkable attack success rate (ASR) of 97.0% and 95.6% in white-box and black-box settings, respectively. These results highlight the urgent need for countermeasures against evasion attacks in Wi-Fi sensing systems. We introduced adversarial training and randomised smoothing, which notably improved the robustness of the Wi-Fi sensing model. The ASRs for white-box and black-box attacks were reduced to a minimum of around 6% and 2%, respectively. Moreover, randomised smoothing also introduced certifiable robustness, achieving 70.1% of samples certified for our model. The certification method provides an additional layer of reliability, ensuring that the model's performance remains consistent and predictable even under adversarial conditions.

Thanks to the ability to customize the propagation of wireless signals, reconfigurable intelligent surfaces (RISs) have great potential in enhancing the performance of future wireless communication systems. While the majority of papers in the literature considers single-RIS scenarios, the potential deployment of multiple RISs, that offer ubiquitous connectivity for diverse user demands, calls for further investigation. This paper considers a downlink multi-group multicast system underpinned by multiple RISs and aims to maximize the sum spectral efficiency subject to an overall transmit power constraint. This optimization problem is highly challenging due to the non-convex, non-smooth, and non-differentiable properties of the objective function, as well as the non-convex unit modulus constraint. To address this complex problem, we propose a model-driven deep learning (DL) approach. This involves first solving the joint active and passive beamforming design through an alternating projected gradient (APG) algorithm with an approximate objective function. The APG algorithm is then unfolded into an iterative procedure using multiple layers with trainable parameters. A network training method is proposed to ensure that the performance improves with the number of iterations. Remarkably, our model is also nicely generalizable to the imperfect channel state information (CSI) scenario, without any change to the network architecture, by simply combining the recursive approximation method and adding some long/short-term trainable parameters to accommodate the two-timescale transmission protocol. Our simulation results demonstrate the superiority of our proposed DL method over existing algorithms in terms of both complexity and performance. Specifically, the proposed model-driven DL method reduces the runtime by approximately 80% compared to the APG algorithm and 99.97% compared to the majorization-minimization algorithm, while it also achieves comparable performance. Furthermore, our proposed method for imperfect CSI scenarios reduces the performance loss by 5%-10% compared to the proposed method without considering the influence of imperfect CSI.

In this paper, we propose a unified framework based on equivariance for the design of artificial intelligence (AI)-assisted technologies in multi-user multiple-input-multiple-output (MU-MIMO) systems. We first provide definitions of multidimensional equivariance, high-order equivariance, and multidimensional invariance (referred to collectively as tensor equivariance). On this basis, by investigating the design of precoding and user scheduling, which are key techniques in MU-MIMO systems, we delve deeper into revealing tensor equivariance of the mappings from channel information to optimal precoding tensors, precoding auxiliary tensors, and scheduling indicators, respectively. To model mappings with tensor equivariance, we propose a series of plug-and-play tensor equivariant neural network (TENN) modules, where the computation involving intricate parameter sharing patterns is transformed into concise tensor operations. Building upon TENN modules, we propose the unified tensor equivariance framework that can be applicable to various communication tasks, based on which we easily accomplish the design of corresponding AI-assisted precoding and user scheduling schemes. Simulation results show that the proposed methods achieve near-optimal performance with significantly lower complexity and strong generalization across multiple dimensions. For instance, the NN trained for precoding with 8 users provides satisfactory performance in a 10-user scenario. This validates the superiority of TENN modules and the unified framework.

Device-to-Device (D2D) link scheduling in wireless communications is a challenging non-convex combinatorial optimization problem. The state-of-the-art methods, either from a model-based or a data-driven perspective, exhibit certain limitations such as the critical need of Channel State Information (CSI) and a large number of instances or solved instances as training samples. To advance this line of research, we propose a novel hybrid model/data-driven approach with Graph Reinforcement Learning for Link Scheduling (GRLinQ), injecting information theoretical insights into machine leaning models. GRLinQ demonstrates superior performance to the existing model-based and data-driven link scheduling mechanisms, with a relaxed requirement of CSI, a smaller number of unsolved instances as training samples, a possible distributed deployment, and more remarkably an excellent generalization ability over different network scenarios and system configurations.

While Graph Neural Networks (GNNs) have achieved enormous success in multiple graph analytical tasks, modern variants mostly rely on the strong inductive bias of homophily. However, real-world networks typically exhibit both homophilic and heterophilic linking patterns, wherein adjacent nodes may share dissimilar attributes and distinct labels. Therefore, GNNs smoothing node proximity holistically may aggregate both task-relevant and irrelevant (even harmful) information, limiting their ability to generalize to heterophilic graphs and potentially causing non-robustness. In this work, we propose a novel Edge Splitting GNN (ES-GNN) framework to adaptively distinguish between graph edges either relevant or irrelevant to learning tasks. This essentially transfers the original graph into two subgraphs with the same node set but complementary edge sets dynamically. Given that, information propagation separately on these subgraphs and edge splitting are alternatively conducted, thus disentangling the task-relevant and irrelevant features. Theoretically, we show that our ES-GNN can be regarded as a solution to a disentangled graph denoising problem, which further illustrates our motivations and interprets the improved generalization beyond homophily. Extensive experiments over 11 benchmark and 1 synthetic datasets not only demonstrate the effective performance of ES-GNN but also highlight its robustness to adversarial graphs and mitigation of the over-smoothing problem.

Graph convolutional networks (GCNs) emerge as the most successful learning models for graph-structured data. Despite their success, existing GCNs usually ignore the entangled latent factors typically arising in real-world graphs, which results in nonexplainable node representations. Even worse, while the emphasis has been placed on local graph information, the global knowledge of the entire graph is lost to a certain extent. In this work, to address these issues, we propose a novel framework for GCNs, termed LGD-GCN, taking advantage of both local and global information for disentangling node representations in the latent space. Specifically, we propose to represent a disentangled latent continuous space with a statistical mixture model, by leveraging neighborhood routing mechanism locally. From the latent space, various new graphs can then be disentangled and learned, to overall reflect the hidden structures with respect to different factors. On the one hand, a novel regularizer is designed to encourage interfactor diversity for model expressivity in the latent space. On the other hand, the factor-specific information is encoded globally via employing a message passing along these new graphs, in order to strengthen intrafactor consistency. Extensive evaluations on both synthetic and five benchmark datasets show that LGD-GCN brings significant performance gains over the recent competitive models in both disentangling and node classification. Particularly, LGD-GCN is able to outperform averagely the disentangled state-of-the-arts by 7.4% on social network datasets.

This paper investigates the joint beam alignment and Doppler estimation (BADE) for fast time-varying wideband millimeter-wave channels, which is essential for subsequent data transmission. In such scenarios, the non-negligible Doppler frequencies significantly impact the beam alignment performance and reference signal overhead, calling for accurate time variation modeling and efficient transceiver design. Toward this end, we leverage the angle, Doppler frequency, and delay sparsity, thus formulating the joint BADE problem as a sparse signal recovery problem in the angle-Doppler-delay domain. The feasibility of the formulated problem strongly depends on the transmitter codebook and the receiver algorithm, which motivates our design. For the transmitter codebook, we characterize the design criterion aiming at maximal identifiable paths, which facilitates precise path parameter estimations and is not satisfied by existing deterministic codebooks. Following this, we provide a new deterministic codebook generation algorithm to meet the necessary conditions of the proposed criterion. For the receiver algorithm, we propose the greedy-based multi-path parameter extraction algorithm. In the proposed algorithm, the hierarchical refinement dictionaries with extended refinement range are employed, balancing the BADE performance and computational complexity. The numerical simulations demonstrate the superiority of the proposed transceiver over benchmarks on the joint BADE problem.

Integrated sensing and communication (ISAC) is emerging as a key enabler to address the increasing demands of spectrum and throughput for ubiquitous sensing and communication. Hereafter, we consider the channel estimation and localization for multi-user ISAC systems assisted by the reconfigurable intelligent surface (RIS) technology. In order to acquire precise environmental information, we propose a novel circular RIS architecture with circularly arranged reflecting unit cells. By modeling the training signal as a low-rank third-order canonical polyadic tensor, we transform the channel estimation problem into a tensor deconstruction task. By leveraging the phase mode excitation principle, we develop a customized RIS training pattern, and retrieve the equivalent channel parameters by subspace estimation algorithms. By exploring the characteristics of RIS array manifolds and free-space propagation, we implement a unique decoupling of channel parameters for user localization, which cannot be supported by traditional linear RIS topologies. Moreover, the design degrees of freedom in the spatial and frequency dimensions are also exploited to further enhance the proposed algorithms. Simulation results indicate that the circular RIS-enabled channel estimation schemes can recover the propagation information with remarkable accuracy, thereby offering a high-level resolution of localization.

In this paper, we consider symbol-level precoding (SLP) in channel-coded multiuser multi-input single-output (MISO) systems. It is observed that the received SLP signals do not always follow Gaussian distribution, rendering the conventional soft demodulation with the Gaussian assumption unsuitable for the coded SLP systems. It, therefore, calls for novel soft demodulator designs for non-Gaussian distributed SLP signals with accurate log-likelihood ratio (LLR) calculation. To this end, we first investigate the non-Gaussian characteristics of both phase-shift keying (PSK) and quadrature amplitude modulation (QAM) received signals with existing SLP schemes and categorize the signals into two distinct types. The first type exhibits an approximate-Gaussian distribution with the outliers extending along the constructive interference region (CIR). In contrast, the second type follows some distribution that significantly deviates from the Gaussian distribution. To obtain accurate LLR, we propose the modified Gaussian soft demodulator and Gaussian mixture model (GMM)-expectation-maximization (EM) soft demodulators to deal with two types of signals respectively. Subsequently, to further reduce the computational complexity and pilot overhead, we put forward a novel neural network named pilot feature extraction network (PFEN) to replace the EM algorithm, leveraging the transformer mechanism in deep learning. Simulation results show that the proposed soft demodulators dramatically improve the throughput of existing SLPs for both PSK and QAM transmission in coded systems.

This paper investigates the robust design of symbol-level precoding (SLP) for multiuser multiple-input multiple-output (MIMO) downlink transmission with imperfect channel state information (CSI) caused by channel aging. By utilizing thea posteriorichannel model based on the widely adopted jointly correlated channel model, the imperfect CSI is modeled as the statistical CSI incorporating the channel mean and channel variance information with spatial correlation. With the signal model in the presence of channel aging, we formulate the signal-to-noise-plus-interference ratio (SINR) balancing and minimum mean square error (MMSE) problems for robust SLP design. The former targets to maximize the minimum SINR across users, while the latter minimizes the mean square error between the received signal and the target constellation point. When it comes to massive MIMO scenarios, the increment in the number of antennas poses a computational complexity challenge, limiting the deployment of SLP schemes. To address such a challenge, we simplify the objective function of the SINR balancing problem and further derive a closed-form SLP scheme. Besides, by approximating the matrix involved in the computation, we modify the proposed algorithm and develop an MMSE-based SLP scheme with lower computation complexity. Simulation results confirm the superiority of the proposed schemes over the state-of-the-art SLP schemes.

Extremely large-scale multiple-input multiple-output (XL-MIMO) constitutes the design trend for base stations of future wireless communication systems, being capable of offering pencil-like beamforming that confronts path loss in an energy-efficient manner. However, wideband wireless applications with XL-MIMO antenna arrays are usually subject to near-field signal propagation conditions, frequency selectivity, and the spatial-wideband effect, whose ignorance in the beamforming optimization process will severely degrade the achievable performance. In this paper, we present an algorithmic framework for designing near-field reception beamforming of wideband multi-user XL-MIMO systems realized with holographic metasurface-based antenna arrays (HMAs). We first present a spherical-wave-propagation channel model, including the near-field effect, frequency selectivity, as well as the spatial-wideband effect. Based on this model, we formulate an HMA-based reception beamforming optimization problem for the uplink of multi-user XL-MIMO communications, whose optimal solution is challenging to obtain due to the nonlinear coupling between the high-dimensional analog combining weights and the digital combiner. To efficiently address the proposed framework via a convergent iterative approach, the considered sum-rate design objective is transformed into a sum-mean-square-error-minimization one. Our extensive numerical investigations showcase that the proposed HMA-based combining scheme can effectively deal with the practical effects under investigation, achieving a higher sum rate than conventional phase-shifter-based hybrid analog and digital combiners having the same antenna aperture.

In recent years, there has been an explosion of research into developing more robust deep neural networks against adversarial examples. Adversarial training appears as one of the most successful methods. To deal with both the robustness against adversarial examples and the accuracy over clean examples, many works develop enhanced adversarial training methods to achieve various trade-offs between them [[19], [38], [80]], Leveraging over the studies [8], [32] that smoothed update on weights during training may help find flat minima and improve generalization, we suggest reconciling the robustness-accuracy trade-off from another perspective, i.e., by adding random noise into deterministic weights. The randomized weights enable our design of a novel adversarial training method via Taylor expansion of a small Gaussian noise, and we show that the new adversarial training method can flatten loss landscape and find flat minima. With PGD, CW, and Auto Attacks, an extensive set of experiments demonstrate that our method enhances the state-of-the-art adversarial training methods, boosting both robustness and clean accuracy. The code is available at https://github.com/Alexkael/Randomized-Adversarial-Training.

This paper investigates the robust design of symbollevel precoding (SLP) for multiuser multiple-input multipleoutput (MIMO) downlink transmission with imperfect channel state information (CSI) caused by channel aging. By utilizing the a posteriori channel model based on the widely adopted jointly correlated channel model, the imperfect CSI is modeled as the statistical CSI incorporating the channel mean and channel variance information with spatial correlation. With the signal model in the presence of channel aging, we formulate the signal-to-noise-plus-interference ratio (SINR) balancing problem for robust SLP design, which targets to maximize the minimum SINR and can be transformed into a typical max-min fractional programming (MMFP). In the scenario of massive MIMO, we simplify the objective function of the SINR balancing problem and further derive a low-complexity SLP scheme. Simulation results confirm the superiority of the proposed schemes over the state-of-the-art SLP schemes.

In this paper, we consider a primal-dual algorithmic aspect of the link scheduling problem in dynamic wireless networks, where the channel characteristics are dramatically time-varying that could render state-of-the-art link scheduling mechanisms computational expensive to fit network dynamics. Building upon the optimality condition of treating interference as noise, we formulate link scheduling as a maximum weighted clique problem on a coexisting graph, which can be transformed into a minimum weighted vertex coloring problem (MWVCP) through the primal-dual technique. With an adversarial perturbation modeling of network dynamics with edge insertion/deletion, we propose dynamic graph algorithms to solve the MWVCP for chordal graph classes with theoretical guarantees on algorithmic feasibility, optimality, and updating complexity. It is expected the resulting link scheduling mechanism could shed light on robust, scalable, and certifiable system designs in dynamic wireless networks from the algorithmic perspective.

The state-of-the-art coding schemes for topological interference management (TIM) problems are usually handcrafted for specific families of network topologies, relying critically on experts' domain knowledge. This inevitably restricts the potential wider applications to wireless communication systems, due to the limited generalizability. This work makes the first attempt to advocate a novel intelligent coding approach to mimic topological interference alignment via local graph coloring algorithms, leveraging the new advances of graph neural networks (GNNs) and reinforcement learning (RL). The extensive experiments demonstrate the excellent generalizability and transferability of the proposed approach, where the parameterized GNNs trained by small size TIM instances are able to work well on new unseen network topologies with larger size.

Channel reciprocity has been commonly assumed in time division duplex (TDD) massive multiple-input multiple-output (MIMO) communications, when acquiring downlink channel state information (CSI) at the base station through the uplink channel estimation. However, such channel reciprocity suffers from severe impairments when low-resolution analog-to-digital converters (ADCs) are introduced to reduce hardware cost and power consumption in practical systems. To compensate for such impairments, in this paper, we propose an efficient calibration state diagnosis scheme built upon compressive sensing techniques, leveraging the sparsity property of the calibration operations at the BS antennas under an additive quantization noise model. Compared with the traditional pilot-based approaches, our proposed scheme achieves comparable accuracy performance with 1-2 quantization bits and hence significantly reduces pilot overhead therein.

Spectral Graph Neural Networks (GNNs) have achieved tremendous success in graph machine learning, with polynomial filters applied for graph convolutions, where all nodes share the identical filter weights to mine their local contexts. Despite the success, existing spectral GNNs usually fail to deal with complex networks (e.g., WWW) due to such homogeneous spectral filtering setting that ignores the regional heterogeneity as typically seen in real-world networks. To tackle this issue, we propose a novel diverse spectral filtering (DSF) framework, which automatically learns node-specific filter weights to exploit the varying local structure properly. Particularly, the diverse filter weights consist of two components — A global one shared among all nodes, and a local one that varies along network edges to reflect node difference arising from distinct graph parts — to balance between local and global information. As such, not only can the global graph characteristics be captured, but also the diverse local patterns can be mined with awareness of different node positions. Interestingly, we formulate a novel optimization problem to assist in learning diverse filters, which also enables us to enhance any spectral GNNs with our DSF framework. We showcase the proposed framework on three state-of-the-arts including GPR-GNN, BernNet, and JacobiConv. Extensive experiments over 10 benchmark datasets demonstrate that our framework can consistently boost model performance by up to 4.92% in node classification tasks, producing diverse filters with enhanced interpretability.

Adversarial training has been shown to be one of the most effective approaches to improve the robustness of deep neural networks. It is formalized as a min-max optimization over model weights and adversarial perturbations, where the weights can be optimized through gradient descent methods like SGD. In this paper, we show that treating model weights as random variables allows for enhancing adversarial training through Second-Order Statistics Optimization (S2O) with respect to the weights. By relaxing a common (but unrealistic) assumption of previous PAC-Bayesian frameworks that all weights are statistically independent, we derive an improved PAC-Bayesian adversarial generalization bound, which suggests that optimizing second-order statistics of weights can effectively tighten the bound. In addition to this theoretical insight, we conduct an extensive set of experiments, which show that S2O not only improves the robustness and generalization of the trained neural networks when used in isolation, but also integrates easily in state-of-the-art adversarial training techniques like TRADES, AWP, MART, and AVMixup, leading to a measurable improvement of these techniques. The code is available at https://github.com/Alexkael/S2O.

To meet the escalating demand on spectral resource in massive machine-type communication (mMTC) applications, a critical solution is applying massive grant-free transmission to the millimeter-wave (mmWave) band. In this paper, to maximize the ergodic rate, we propose an efficient hybrid analog/digital beamforming (HBF) design algorithm for the massive grant-free transmission in uplink mmWave systems. Specifically, to make the HBF design problem tractable, we first leverage the deterministic equivalent method to derive an approximate expression of the ergodic rate for the mMTC in the mmWave system. Since the ergodic rate maximization-based HBF design problem is nonconvex, we leverage the alternating optimization strategy and propose a semidefinite relaxation-based HBF algorithm to improve the ergodic rate. Simulation results verify the superior performance of the proposed HBF design algorithm in improving the ergodic rate.

In this paper, we consider the topological interference management (TIM) problem in a dynamic setting, where an adversary perturbs network topology to prevent the exploitation of sophisticated coding opportunities (e.g., interference alignment). Focusing on a special class of network topology – chordal networks – we investigate algorithmic aspects of the TIM problem under adversarial topology perturbation. In particular, given the adversarial perturbation with respect to edge insertion/deletion, we propose a dynamic graph coloring algorithm that allows for a constant number of re-coloring updates against each inserted/deleted edge to achieve the information-theoretic optimality. This is a sharp reduction of the general graph re-coloring, whose optimal number of updates scales as the size of the network, thanks to the delicate exploitation of the structural properties of chordal graph classes.

In the massive grant-free orthogonal frequency division multiple access (OFDMA), the timing and frequency offsets between users impose new challenges on joint active user detection (AUD) and channel estimation (CE) for the subsequent data recovery. In the asynchronous OFDMA, the timing and frequency offset effects can be modeled as the phase-shifting on the pilot matrix. As such, by constructing the measurement matrix with timing and frequency offsets, the joint estimation problem can be formulated as a multiple measurement vector (MMV) recovery problem with structured sparsity. However, such structured sparsity cannot be tackled by the existing compressed sensing (CS) techniques. To address this issue, we develop an efficient structured generalized approximate message passing (S-GAMP) algorithm, which includes the parallel AMP-MMV algorithm as a particular case. To deal with the high dimensionality of the measurement matrix, we propose the dynamic S-GAMP algorithm with a dynamic measurement matrix to reduce the computational complexity. Simulation results confirm the superiority of the proposed algorithms in grant-free OFDMA with both timing and frequency offsets.

We consider the pilot assignment problem in large-scale distributed multi-input multi-output (MIMO) networks, where a large number of remote radio head (RRH) antennas are randomly distributed in a wide area, and jointly serve a relatively smaller number of users (UE) coherently. By artificially imposing structures on the UE-RRH connectivity, we model the network by a partially-connected interference network, so that the pilot assignment problem can be cast as a topological interference management problem with multiple groupcast messages. Building upon such connection, we formulate the topological pilot assignment (TPA) problem in two different ways with respect to whether or not the to-be-estimated channel connectivity pattern is knowna priori. When it is known, we formulate the TPA problem as a low-rank matrix completion problem that can be solved by a simple alternating projection algorithm. Otherwise, we formulate it as a sequential maximum weight induced matching problem that can be solved by either a mixed integer linear program or a simple yet efficient greedy algorithm. With respect to two different formulations of the TPA problem, we evaluate the efficiency of the proposed algorithms under the cell-free massive MIMO setting.

In this paper, we consider user selection and downlink precoding for an over-loaded single-cell massive multiple-input multiple-output (MIMO) system in frequency division duplexing (FDD) mode, where the base station is equipped with a dual-polarized uniform planar array (DP-UPA) and serves a large number of single-antenna users. Due to the absence of uplink-downlink channel reciprocity and the high-dimensionality of channel matrices, it is extremely challenging to design downlink precoders using closed-loop channel probing and feedback with limited spectrum resource. To address these issues, a novel methodology – active channel sparsification (ACS) – has been proposed recently in the literature for uniform linear array (ULA) to design sparsifying precoders, which substantially reduces channel feedback overhead. Pushing forward this line of research, we aim to facilitate the potential deployment of ACS in practical FDD massive MIMO systems, by extending it from ULA to DP-UPA with explicit user selection and making the current ACS implementation simplified. To this end, by leveraging Toeplitz matrix theory, we start with the spectral properties of channel covariance matrices from the lens of their matrix-valued spectral density function. Inspired by these properties, we extend the original ACS using scalar-weight bipartite graph representation to the matrix-weight counterpart. Building upon such matrix-weight bipartite graph representation, we propose a multi-dimensional ACS (MD-ACS) method, which is a generalization of original ACS formulation and is more suitable for DP-UPA antenna configurations. The nonlinear integer program formulation of MD-ACS can be classified as a generalized multi-assignment problem (GMAP), for which we propose a simple yet efficient greedy algorithm to solve it. Simulation results demonstrate the performance improvement of the proposed MD-ACS with greedy algorithm over the state-of-the-art methods based on the QuaDRiGa channel models.

This tutorial aims to introduce the fundamentals of adversarial robustness of deep learning, presenting a well-structured review of up-to-date techniques to assess the vulnerability of various types of deep learning models to adversarial examples. This tutorial will particularly highlight state-of-the-art techniques in adversarial attacks and robustness verification of deep neural networks (DNNs). We will also introduce some effective countermeasures to improve robustness of deep learning models, with a particular focus on adversarial training. We aim to provide a comprehensive overall picture about this emerging direction and enable the community to be aware of the urgency and importance of designing robust deep learning models in safety-critical data analytical applications, ultimately enabling the end-users to trust deep learning classifiers. We will also summarize potential research directions concerning the adversarial robustness of deep learning, and its potential benefits to enable accountable and trustworthy deep learning-based data analytical systems and applications.

We consider the pilot assignment problem in cell-free massive multi-input multi-output (MIMO) networks, where a large number of remote radio head (RRH) antennas are randomly distributed in a wide area, and jointly serve a relatively smaller number of users (UE) coherently. By artificially imposing topological structures on the UE-RRH connectivity, we model the network by a partially-connected interference network and formulate the topological pilot assignment (TPA) problem as a sequential maximum weight induced matching problem that can be solved by either a mixed integer linear program or a simple yet efficient greedy algorithm. The efficiency of the proposed algorithms is evaluated in cell-free massive MIMO networks.

In this paper, we consider massive multiple-input multiple-output (MIMO) communication systems with a uniform planar array (UPA) at the base station (BS) and investigate the downlink precoding with imperfect channel state information (CSI). By exploiting both instantaneous and statistical CSI, we aim to design precoding vectors to maximize the ergodic rate subject to a total transmit power constraint. By maximizing an upper bound of the ergodic rate instead, we leverage the corresponding Lagrangian formulation and identify the structural characteristics of the optimal precoder as the solution to a generalized eigenvalue problem. As such, the high-dimensional precoder design problem turns into a low-dimensional power control problem. The Lagrange multipliers play a crucial role in determining both precoder directions and power parameters, yet are challenging to be solved directly. To figure out the Lagrange multipliers, we develop a deep learning approach underpinned by a properly designed neural network that learns directly from CSI. With the offline pre-trained neural network, the online computational complexity of precoding is substantially reduced compared with the existing iterative algorithm while maintaining nearly the same performance.

While adversarial training becomes one of the most promising defending approaches against adversarial attacks for deep neural networks, the conventional wisdom through robust optimization may usually not guarantee good generalization for robustness. Concerning with robust generalization over unseen adversarial data, this paper investigates adversarial training from a novel perspective of shift consistency in latent space. We argue that the poor robust generalization of adversarial training is owing to the significantly dispersed latent representations generated by training and test adversarial data, as the adversarial perturbations push the latent features of natural examples in the same class towards diverse directions. This is underpinned by the theoretical analysis of the robust generalization gap, which is upper-bounded by the standard one over the natural data and a term of feature inconsistent shift caused by adversarial perturbation {–} a measure of latent dispersion. Towards better robust generalization, we propose a new regularization method {–} shift consistency regularization (SCR) {–} to steer the same-class latent features of both natural and adversarial data into a common direction during adversarial training. The effectiveness of SCR in adversarial training is evaluated through extensive experiments over different datasets, such as CIFAR-10, CIFAR-100, and SVHN, against several competitive methods.

In this paper, we consider the topological interference management (TIM) problem in a dynamic setting, where an adversary perturbs network topology to prevent the exploitation of sophisticated coding opportunities (e.g., interference alignment). Focusing on a special class of network topology - chordal networks - we investigate algorithmic aspects of the TIM problem under adversarial topology perturbation. In particular, given the adversarial perturbation with respect to edge insertion/deletion, we propose a dynamic graph coloring algorithm that allows for a constant number of re-coloring updates against each inserted/deleted edge to achieve the information-theoretic optimality. This is a sharp reduction of the general graph re-coloring, whose optimal number of updates scales as the size of the network, thanks to the delicate exploitation of the structural properties of chordal graph classes.

While massive multiple-input multiple-output (MIMO) has achieved tremendous success in both theory and practice, it faces a crisis of sharp performance degradation in moderate or high-mobility scenarios (e.g., 30 km/h), due to the breach of uplink-downlink channel duality. Such a “curse of mobility” has spurred the research on channel prediction in high-mobility scenarios. Instead of predicting channel response matrix in the space-frequency domain, we investigate it in the angle-delay domain by utilizing the high angle-delay resolution of wideband massive MIMO systems. Specifically, we study the general angle-delay domain channel characterization and obtain that: 1) the correlations between the angle-delay domain channel response matrix (ADCRM) elements are decoupled significantly; 2) when the number of antennas and bandwidth are limited, the decoupling is insufficient and residual correlations between the neighboring ADCRM elements exist. Then focusing on the ADCRM, we propose two channel prediction methods: a spatio-temporal autoregressive (ST-AR) model-driven unsupervised-learning method and a deep learning (DL) based data-driven supervised-learning method. While the model-driven method provides a principled way for channel prediction, the data-driven method is generalizable to various channel scenarios. In particular, ST-AR exploits the residual spatio-temporal correlations of the channel element with its most neighboring elements, and DL realizes element-wise angle-delay domain channel prediction utilizing a complex-valued neural network (CVNN). Simulation results under the 3GPP non-line-of-sight (NLOS) scenarios indicate that, compared to the state-of-the-art Prony-based angular-delay domain (PAD) prediction method, both the proposed ST-AR and the CVNN-based channel prediction methods can enhance the channel prediction accuracy.

In this paper, we consider massive multiple-input-multiple-output (MIMO) communication systems with a uniform planar array (UPA) at the base station (BS) and investigate the downlink precoder design with imperfect channel state information (CSI). By exploiting channel estimates and statistical parameters of channel estimation error, we aim to design precoding vectors to maximize the utility function on the ergodic rates of users subject to a total transmit power constraint. By employing an upper bound of the ergodic rate, we leverage the corresponding Lagrangian formulation and identify the structural characteristics of the optimal precoder as the solution to a generalized eigenvalue problem. The Lagrange multipliers play a crucial role in determining both precoding directions and power parameters, yet are challenging to be solved directly. To figure out the Lagrange multipliers, we develop a general framework underpinned by a properly designed neural network that learns directly from CSI. To further relieve the computational burden, we obtain a low-complexity framework by decomposing the original problem into computationally efficient subproblems with instantaneous and statistical CSI handled separately. With the offline pre-trained neural network, the online computational complexity of precoder is substantially reduced compared with the existing iterative algorithm while maintaining nearly the same performance.

In this article, we investigate multi-cell scheduling for massive multiple-input-multiple-output (MIMO) communications with only statistical channel state information (CSI). The objective of multi-cell scheduling is to activate a subset of users so as to maximize the ergodic sum rate subject to per-cell total transmit power constraint. By adopting beam division multiple access based on the statistical CSI, i.e., channel-coupling matrix (CCM), we simplify multi-cell scheduling as a power control problem in the beam domain, by which the ergodic sum rate is maximized. To reduce the computational burden on finding the ergodic sum rate, we propose a learning-to-compute strategy, which directly computes the complex ergodic rate function from CCMs via a deep neural network. Specifically, by modeling the probability density function of the ordered eigenvalues of the Hermitian CCM matrices as exponential family distributions, a properly designed hybrid neural network makes the ergodic rate computation feasible. With the learning-to-compute strategy, the online computational complexity of multi-cell scheduling is substantially reduced compared with the existing Monte Carlo or deterministic equivalent (DE) based methods while maintaining nearly the same performance.

In this paper, we investigate user positioning in massive multiple-input multiple-output (MIMO) orthogonal frequency-division multiplexing (OFDM) systems where the base station (BS) is equipped with a uniform planar array (UPA). Taking advantage of the UPA geometry and wide bandwidth, we advocate the use of the angle-delay channel power matrix (ADCPM) as a new type of fingerprint to replace the traditional ones. The ADCPM embeds the stable and stationary multipath characteristics, e.g., delay, power, and angles in the vertical and horizontal directions, which are beneficial to positioning. We further exploit the sparsity of the ADCPM to reduce the noise contamination in the ADCPM. Taking ADCPM fingerprints as the inputs, we propose a novel three-dimensional (3D) convolution neural network (CNN) enabled learning method to localize the 3D positions of the mobile terminals (MTs). In particular, such a 3D CNN model consists of a convolution refinement module to refine the elementary feature maps from the ADCPM fingerprints, three extended Inception modules to extract the advanced feature maps, and a regression module to estimate the 3D positions. By intensive simulations, the proposed 3D CNN-enabled positioning method is demonstrated to achieve higher positioning accuracy than the traditional searching-based ones, with reduced computational complexity and storage overhead, and robust to noise contamination.

We consider a single-cell massive multi-input multi-output (MIMO) network with uniform planar array (UPA) antennas equipped at the base station that serves a number of single-antenna users. In the overloaded multi-user setting, it is likely that users' channels are highly spatial-correlated with overlapping spectrum in the angular domain, which imposes challenges on uplink channel estimation and data transmission due to potential pilot contamination during uplink training and multiuser interference during uplink data transmission. To mitigate the effect of multiuser channel spatial correlation, we adopt a recently proposed active channel sparsification strategy, and propose a novel method for joint user and beam selection in the angular domain. In particular, we represent all users' channels in the angular/beam domain, taking advantage of the doubly block Toeplitz structure of the channel covariance matrix for UPA. Accordingly, we construct a weighted bipartite graph to represent the beam and user association for ease of user/beam selection. By doing so, we reformulate the problems of mean square error minimization for uplink channel estimation and sum rate maximization for uplink data detection as two mixed integer linear programs (MILPs), by which the challenging joint user and beam selection problem can be efficiently solved via off-the-shelf MILP solvers. The simulation results demonstrate the effectiveness of our active channel sparsification strategy for the joint user and beam selection.

In this paper, we investigate the three-dimensional (3D) user positioning in massive multiple-input multiple-output (MIMO) orthogonal frequency-division multiplexing (OFDM) systems with the base station (BS) equipped with a uniform planner antenna (UPA) array. Taking advantage of the UPA array geometry and wide bandwidth, we advocate the use of the angle-delay channel power matrix (ADCPM) as a new type of fingerprint to replace the traditional ones. The ADCPM embeds the stable and stationary multipath characteristics, e.g., delay, power, and angles in the vertical and horizontal directions, which are beneficial to positioning. Taking ADCPM fingerprints as the inputs, we propose a novel 3D convolution neural network (CNN) enabled learning method to localize users' 3D positions. By intensive simulations, the proposed 3D CNN-enabled positioning method is demonstrated to achieve higher positioning accuracy than the traditional searching-based ones, with reduced computational complexity and storage overhead, and robust to noise contamination.

This paper studies the novel concept of weight correlation in deep neural networks and discusses its impact on the networks' generalisation ability. For fully-connected layers, the weight correlation is defined as the average cosine similarity between weight vectors of neurons, and for convolutional layers, the weight correlation is defined as the cosine similarity between filter matrices. Theoretically, we show that, weight correlation can, and should, be incorporated into the PAC Bayesian framework for the generalisation of neural networks, and the resulting generalisation bound is monotonic with respect to the weight correlation. We formulate a new complexity measure, which lifts the PAC Bayes measure with weight correlation, and experimentally confirm that it is able to rank the generalisation errors of a set of networks more precisely than existing measures. More importantly, we develop a new regulariser for training, and provide extensive experiments that show that the generalisation error can be greatly reduced with our novel approach.

The topological interference management (TIM) problem studies the degrees of freedom (DoF) of partially-connected interference networks with no channel state information (CSI) at the transmitters except the network topology (i.e., partial connectivity). In this paper, we consider a variant of the TIM problem with uncertainty in network topology, where the channel state with partial connectivity is only known to belong to one of M states at the transmitters. In particular, the transmitter has access to all network topological information over M states, but is unaware of which state it falls in exactly for communication. The receiver at any state is aware of the exact state it falls in besides the network topologies of all states, and wish to recover as much highly-prioritized information at current state as possible. We formulate it as the opportunistic TIM problem with network uncertainty modeled by M state-varying network topologies. To adapt to network topology uncertainty and different message decoding priority, joint encoding and opportunistic decoding are enabled at the transmitters and receivers respectively. Specifically, being aware of all possible network topologies, each transmitter sends a signal jointly encoded from all messages desired over M states, say M distinct messages, and at a certain State m, Receiver k wishes to opportunistically decode the first πk(m) ∈ {1, 2, · · · , M} higher-priority messages. Under this opportunistic TIM setting, we construct a multi-state conflict graph to capture the mutual conflict of messages over M states, and characterize the optimal DoF region of two classes of network topologies via polyhedral combinatorics. A remarkable fact is that, under an additional mild monotonous condition, the optimality conditions of orthogonal access and one-to-one interference alignment still apply to TIM with uncertainty in network topology.

We consider the information-theoretic optimality of treating inter-cell interference as noise (multi-cell TIN) in downlink cellular networks. We focus on scenarios modeled by the Gaussian interfering broadcast channel (IBC), comprising K mutually interfering Gaussian broadcast channels (BCs), each formed by a base station communicating independent messages to an arbitrary number of users. We establish a new power allocation duality between the IBC and its dual interfering multiple access channel (IMAC), which entails that the corresponding generalized degrees-of-freedom regions achieved through multi-cell TIN and power control (TINA regions) for both networks are identical. As by-products of this duality, we obtain an explicit characterization of the IBC TINA region from a previously established characterization of the IMAC TINA region; and identify a multi-cell convex-TIN regime in which the IBC TINA region is a polyhedron (hence convex) without the need for time-sharing. We then identify a smaller multi-cell TIN regime in which the IBC TINA region is optimal and multi-cell TIN achieves the entire capacity region of the IBC, up to a constant gap. This is accomplished by deriving a new genie-aided outer bound for the IBC, that reveals a novel BC-type order that holds amongst users in each constituent BC (or cell) under inter-cell interference, which in turn is not implied by previously known BC-type orders (i.e. degraded, less noisy and more capable orders). The multi-cell TIN regime that we identify for the IBC coincides with a corresponding multi-cell TIN regime previously identified for the IMAC, hence establishing a comprehensive uplink-downlink duality of multi-cell TIN in the GDoF (and approximate capacity) sense.

We consider a K-user interference network with M states, where each transmitter has up to M messages and over State m, Receiver k wishes to decode the first πk(m) ∈ {1,2, ⋯, M} messages from its desired transmitter. This problem of channel with states models opportunistic communications, where more messages are decoded for better channel states. The first message from each transmitter has the highest priority as it is required to be decoded regardless of the state of the receiver; the second message is opportunistically decoded if the state allows a receiver to decode 2 messages; and the M-th message has the lowest priority as it is decoded if and only if the receiver wishes to decode all M messages. For this interference network with states, we show that if any possible combination of the channel states satisfies a condition under which power control and treating interference as noise (TIN) are sufficient to achieve the entire generalized degrees of freedom (GDoF) region of this channel state by itself, then a simple layered superposition encoding scheme with power control and a successive decoding scheme with TIN achieves the entire GDoF region of the network with M states for all K M messages.

We investigate energy efficiency (EE) optimization for single-cell massive multiple-input multiple-output (MIMO) downlink transmission with only statistical channel state information (CSI) available at the base station. We first show that beam domain transmission is favorable for energy efficiency in the massive MIMO downlink, by deriving a closed-form solution for the eigenvectors of the optimal transmit covariance matrix. With this conclusion, the EE optimization problem is reduced to a real-valued power allocation problem, which is much easier to tackle than the original large-dimensional complex matrix-valued precoding design problem. We further propose an iterative water-filling-structured beam domain power allocation algorithm with low complexity and guaranteed convergence, exploiting the techniques from sequential optimization, fractional optimization, and random matrix theory. Numerical results demonstrate the near-optimal performance of our proposed statistical CSI aided EE optimization approach.

We investigate energy efficiency (EE) optimization for massive multiple-input multiple-output (MIMO) transmission in a single cell downlink scenario where the base station has only access to statistical channel state information (CSI) of the user terminals. To maximize the system EE, we first figure out a solution for the eigenvectors of the optimal transmit covariance matrices in a closed form. Notably, such a solution indicates that it is more favorable to perform energy efficient transmission in the beam domain for massive MIMO downlink, by which we reformulate the original complicated EE optimization precoding design to a simpler power allocation problem in the beam domain. Exploiting the approaches of sequential optimization, fractional optimization, and deterministic equivalent, we further propose an iterative algorithm for power allocation in the beam domain with guaranteed convergence to a stationary point. Numerical results demonstrate the superior performance and the fast convergence of our proposed statistical CSI aided EE optimization approach for massive MIMO downlink.

We consider the information-theoretic optimality of treating inter-cell interference as noise in downlink cellular networks modeled as Gaussian interfering broadcast channels. Establishing a new uplink-downlink duality, we cast the problem in Gaussian interfering broadcast channels to that in Gaussian interfering multiple access channels, and characterize an achievable GDoF region under power control and treating inter-cell interference as (Gaussian) noise. We then identify conditions under which this achievable GDoF region is optimal.

The topological interference management (TIM) problem studies the degrees of freedom (DoF) of partially- connected interference networks with no channel state information (CSI) at the transmitters except the network topology (i.e., partial connectivity). In this paper, we consider a variant of the TIM problem with uncertainty in network topology, where the channel state with partial connectivity is only known to belong to one of M states at the transmitters. In particular, the transmitter has access to all network topological information over M states, but is unaware of which state it falls in exactly for communication. The receiver at any state is aware of the exact state it falls in besides the network topologies of all states, and wish to recover as much highly-prioritized information at current state as possible. We formulate it as the opportunistic TIM problem with network uncertainty modeled by M state-varying network topologies. To adapt to network topology uncertainty and different message decoding priority, joint encoding and opportunistic decoding are enabled at the transmitters and receivers respectively. Specifically, being aware of all possible network topologies, each transmitter sends a signal jointly encoded from all messages desired over M states, say M distinct messages, and at a certain State m, Receiver k wishes to opportunistically decode the first πk(m)∈ {1,2,...,M} higher-priority messages. Under this opportunistic TIM setting, we construct a multi-state conflict graph to capture the mutual conflict of messages over M states, and characterize the optimal DoF region of two classes of network topologies via polyhedral combinatorics. A remarkable fact is that, under an additional mild monotonous condition, the optimality conditions of orthogonal access and one-to-one interference alignment still apply to TIM with uncertainty in network topology.

We propose a novel method for massive multiple-input multiple-output (massive MIMO) in frequency division duplexing (FDD) systems. Due to the large frequency separation between uplink (UL) and downlink (DL) in FDD systems, channel reciprocity does not hold. Hence, in order to provide DL channel state information to the base station (BS), closed-loop DL channel probing, and channel state information (CSI) feedback is needed. In massive MIMO, this typically incurs a large training overhead. For example, in a typical configuration with M ≃200 BS antennas and fading coherence block of T ≃ 200 symbols, the resulting rate penalty factor due to the DL training overhead, given by max{0, 1 - M/T }, is close to 0. To reduce this overhead, we build upon the well-known fact that the angular scattering function of the user channels is invariant over frequency intervals whose size is small with respect to the carrier frequency (as in current FDD cellular standards). This allows us to estimate the users' DL channel covariance matrix from UL pilots without additional overhead. Based on this covariance information, we propose a novel sparsifying precoder in order to maximize the rank of the effective sparsified channel matrix subject to the condition that each effective user channel has sparsity not larger than some desired DL pilot dimension Tdl, resulting in the DL training overhead factor max{0, 1 - Tdl/T } and CSI feedback cost of Tdl pilot measurements. The optimization of the sparsifying precoder is formulated as a mixed integer linear program, that can be efficiently solved. Extensive simulation results demonstrate the superiority of the proposed approach with respect to the concurrent state-of-the-art schemes based on compressed sensing or UL/DL dictionary learning.

In this paper, we propose a novel method for efficient implementation of a massive Multiple-Input Multiple- Output (massive MIMO) system with Frequency Division Duplexing (FDD) operation. Our main objective is to reduce the large overhead incurred by Downlink (DL) common training and Uplink (UL) feedback needed to obtain channel state information (CSI) at the base station. Our proposed scheme relies on the fact that the underlying angular distribution of a channel vector, also known as the angular scattering function, is a frequency-invariant entity yielding a ULDL reciprocity and has a limited angular support. We estimate this support from UL CSI and interpolate it to obtain the corresponding angular support of the DL channel. Finally we exploit the estimated support of the DL channel of all the users to design an efficient channel probing and feedback scheme that maximizes the total spectral efficiency of the system. Our method is different from the existing compressed-sensing (CS) based techniques in the literature. Using support information helps reduce the feedback overhead from O(s logM) in CS techniques to O(s) in our proposed method, with s andM being sparsity order of the channel vectors and the number of base station antennas, respectively. Furthermore, in order to control the channel sparsity and therefore the DL common training and UL feedback overhead, we introduce the novel concept of active channel sparsification. In brief, when the fixed pilot dimension is less than the required amount for reliable channel estimation, we introduce a pre-beamforming matrix that artificially reduces the effective channel dimension of each user to be not larger than the DL pilot dimension, while maximizing both the number of served users and the number of probed angles. We provide numerical experiments to assess the performance of our method and compare it with the state-of-the-art CS technique.

The topological interference management (TIM) problem studies partially-connected interference networks with no channel state information except for the network topology (i.e., connectivity graph) at the transmitters. In this paper, we consider a similar problem in the uplink cellular networks, while message passing is enabled at the receivers (e.g., base stations), so that the decoded messages can be routed to other receivers via backhaul links to help further improve network performance. For this TIM problem with decoded message passing (TIM-MP), we model the interference pattern by conflict digraphs, connect orthogonal access to the acyclic set coloring on conflict digraphs, and show that one-to-one interference alignment boils down to orthogonal access because of message passing. With the aid of polyhedral combinatorics, we identify the structural properties of certain classes of network topologies where orthogonal access achieves the optimal degrees-of-freedom (DoF) region in the information-theoretic sense. The relation to the conventional index coding with simultaneous decoding is also investigated by formulating a generalized index coding problem with successive decoding as a result of decoded message passing. The properties of reducibility and criticality are also studied, by which we are able to prove the linear optimality of orthogonal access in terms of symmetric DoF for the networks up to four users with all possible network topologies (218 instances). Practical issues of the tradeoff between the overhead of message passing and the achievable symmetric DoF are also discussed, in the hope of facilitating efficient backhaul utilization.

The main result of this paper is that an orthogonal access scheme, such as time division multiple access achieves the all-unicast degrees of freedom (DoF) region of the topological interference management problem if and only if the network topology graph is chordal bipartite, i.e., every cycle that can contain a chord, does contain a chord. The all-unicast DoF region includes the DoF region for any arbitrary choice of a unicast message set, so e.g., the results of Maleki and Jafar on the optimality of orthogonal access for the sum-DoF of one-dimensional convex networks are recovered as a special case. The result is also established for the corresponding topological representation of the index coding problem.

We study the topological interference management problem with decoded message passing (TIM-MP) using a polyhedral approach. The TIM-MP problem considers partially-connected interference channels with no channel state information except for the network topology (i.e., connectivity graph) at the transmitters, while the decoded messages at the receivers can be routed to other receivers via backhaul links to help cancel interference. With the aid of polyhedral combinatorics, we identify the structural properties of certain classes of network topologies for which orthogonal access achieves the optimal degrees-of-freedom (DoF) region in the information-theoretic sense. We are also able to prove the linear optimality of orthogonal access in terms of symmetric DoF for the networks up to four users with all possible network topologies (218 instances).

Topological interference management (TIM) problem studies partially connected interference networks with no channel state information except for the connectivity graph at transmitters. In this paper, we consider a similar problem in the uplink cellular networks while message passing is enabled at receivers (e.g., base stations) in which the decoded messages can be routed to other receivers via backhaul links to help improve overall network performance. For this new problem setting, we try to answer the following two questions: (1) when is orthogonal access optimal? and (2) when does message passing help? From both graph theoretic and index coding perspectives, we are able to offer preliminary answers to those questions by identifying sufficient and/or necessary conditions.

Cache-aided network architectures are emerging as an innovative solution able to harness device memory, a cheap and widely available resource, into bandwidth, so as to meet the predicted dramatic increase of user data traffic generated by on-demand multi-media. In this paper, starting from the previously proposed and widely studied femtocaching network, we consider a partially connected interference network where the femto base stations are equipped with caches and have no access to channel state information beyond the network connectivity (network topology). We aim at characterizing the tradeoff between the cache memory size and the normalized transmission delay for file delivery. We formulate a joint file placement and delivery optimization problem, and propose approaches to compute extreme points of the achievable memory-delay region. Our algorithmic solution consists of decomposing the intractable joint optimization problem into separate subproblems, which are solvable using existing efficient methods.

For single-antenna Gaussian interference channels, we reformulate the problem of determining the generalized degrees of freedom (GDoF) region achievable by treating interference as Gaussian noise (TIN) derived by Geng et al. from a combinatorial optimization perspective. We show that the TIN power control problem can be cast into an assignment problem, such that the globally optimal power allocation variables can be obtained by well-known polynomial time algorithms (e.g., centralized Hungarian method or distributed Auction algorithm). Furthermore, the expression of the TIN-achievable GDoF region (TINA region) can be substantially simplified with the aid of maximum weighted matchings. We also provide conditions under which the TINA region is a convex polytope that relax those by Geng et al. For these new conditions, together with a channel connectivity (i.e., interference topology) condition, we show TIN optimality for a new class of interference networks that is not included, nor includes, the class found by Geng et al. Building on the above insights, we consider the problem of joint link scheduling and power control in wireless networks, which has been widely studied as a basic physical layer mechanism for device-to-device communications. Inspired by the relaxed TIN channel strength condition as well as the assignment-based power allocation, we propose a low-complexity GDoF-based distributed link scheduling and power control mechanism (ITLinQ+) that improves upon the ITLinQ scheme proposed by Naderializadeh and Avestimehr and further improves over the heuristic approach known as FlashLinQ. It is demonstrated by simulation that ITLinQ+ without power control provides significant average network throughput gains over both ITLinQ and FlashLinQ, and yet still maintains the same level of implementation complexity. Furthermore, when ITLinQ+ is augmented by power control, it provides an energy efficiency substantially larger than that of ITLinQ and FlashLinQ, at the cost of additional complexity and some signaling overhead.

For single-antenna Gaussian interference channels, we re-formulate the problem of determining the Generalized Degrees of Freedom (GDoF) region achievable by treating interference as noise (TIN) with proper power control from a combinatorial optimization perspective. We show that the TIN power control problem can be cast into an assignment problem, such that the globally optimal power allocation variables can be obtained by well-known polynomial time algorithms. Furthermore, the expression of the TIN-achievable GDoF region can be substantially simplified with the aid of maximum weighted matchings. In addition, we provide conditions under which the TIN-achievable GDoF region is a convex polytope that relax those in [1]. For these new conditions, together with a channel connectivity (i.e., interference topology) condition, we can prove GDoF optimality for a new class of interference networks that is not included, nor includes, the class found in [1].

Interference networks with no channel state information at the transmitter except for the knowledge of the connectivity graph have been recently studied under the topological interference management framework. In this paper, we consider a similar problem with topological knowledge but in a distributed broadcast channel setting, i.e., a network where transmitter cooperation is enabled. We show that the topological information can also be exploited in this case to strictly improve the degrees of freedom (DoF) as long as the network is not fully connected, which is a reasonable assumption in practice. Achievability schemes from graph theoretic and interference alignment perspectives are proposed. Together with outer bounds built upon generator sequence, the concept of compound channel settings, and the relation to index coding, we characterize the symmetric DoF for the so-called regular networks with constant number of interfering links, and identify the sufficient and/or necessary conditions for the arbitrary network topologies to achieve a certain amount of symmetric DoF.

Interference networks with no channel state information at the transmitter (CSIT) except for the knowledge of the connectivity graph have been recently studied under the topological interference management (TIM) framework. In this paper, we consider a similar topological knowledge but in a distributed broadcast channel setting, i.e. a network where transmitter cooperation is enabled. We show that the interference topology can also be exploited in this case to strictly improve the degrees of freedom (DoF) as long as the network is not fully connected, which is a reasonable assumption in practice. A fractional graph coloring based interference avoidance and a subspace interference alignment approaches are proposed to characterize the symmetric DoF for so-called regular networks with constant interfering degree, and to identify achievable DoF for arbitrary network topologies.

We consider the temporally correlated multiple-input multiple-output (MIMO) broadcast channels (BC) and interference channels (IC) where the transmitter(s) has/have 1) delayed channel state information (CSI) obtained from a latency-prone feedback channel as well as 2) imperfect current CSIT, obtained, e.g., from prediction on the basis of these past channel samples based on the temporal correlation. The degrees of freedom (DoF) regions for the two-user broadcast and interference MIMO networks with general antenna configuration under such conditions are fully characterized, as a function of the prediction quality indicator. Specifically, a simple unified framework is proposed, allowing us to attain optimal DoF region for the general antenna configurations and current CSIT qualities. Such a framework builds upon block-Markov encoding with interference quantization, optimally combining the use of both outdated and instantaneous CSIT. A striking feature of our work is that, by varying the power allocation, every point in the DoF region can be achieved with one single scheme. As a result, instead of checking the achievability of every corner point of the outer bound region, as typically done in the literature, we propose a new systematic way to prove the achievability.

We consider the problem of downlink precoding for Network (multi-cell) MIMO networks where Transmitters (TXs) are provided with imperfect Channel State Information (CSI). Specifically, each TX receives a delayed channel estimate with the delay being specific to each channel component. This model is particularly adapted to the scenarios where a user feeds back its CSI to its serving base only as it is envisioned in future LTE networks. We analyze the impact of the delay during the backhaul-based CSI exchange on the rate performance achieved by Network MIMO. We highlight how delay can dramatically degrade system performance if existing precoding methods are to be used. We propose an alternative robust beamforming strategy which achieves the maximal performance, in DoF sense. We verify by simulations that the theoretical DoF improvement translates into a performance increase at finite Signal-to-Noise Ratio (SNR) as well1.

The degrees of freedom (DoF) of a K-User MISO broadcast channel (BC) is studied when the transmitter (TX) has access to a delayed channel estimate in addition to an imperfect estimate of the current channel. The current estimate could be for example obtained from prediction applied on past estimates, in the case where feedback delay is within the coherence time. Prior results in this setting are promising, yet remain limited to the two-user case. In contrast, we consider here an arbitrary number of users. We develop a new transmission scheme, called the Kα-MAT scheme, which builds upon both the principle of the MAT alignment from Maddah-Ali and Tse and zero-forcing (ZF) to achieve a larger DoF in the channel state information (CSI) configuration previously described. We also develop a new upper bound for the DoF to compare with the DoF achieved by Kα-MAT. Although not optimal, the Kα-MAT scheme performs well when the CSIT quality is not too delayed or K is small. The Kα-MAT scheme can be seen as a robust version of ZF with respect to the delay in the CSI feedback.

The two-user Multiple-Input Multiple-Output (MIMO) broadcast channel (BC) with arbitrary antenna configuration is considered, in which the transmitter obtains (i) delayed channel state information (CSI) from a latency-prone feedback channel as well as (ii) imperfect current CSI, e.g., from prediction based on these past channel samples. The degrees of freedom (DoF) region under such a setting is fully characterized as a function of a prediction quality exponent. This work extends prior work, previously limited to MISO, to fully general antenna settings. An intriguing by-product of our results is to reveal the benefits of dealing with an asymmetric multi-user MIMO configuration (i.e., one in which terminals do not have the same number of antennas) in the case of non-perfect CSIT (e.g., caused by feedback delays or limited preciseness).

We consider the time correlated multiple-input single-output (MISO) broadcast channel where the transmitter has imperfect knowledge of the current channel state, in addition to delayed channel state information. By representing the quality of the current channel state information asP-αfor the signal-to-noise ratioPand some constant α ≥ 0, we characterize the optimal degrees of freedom region for this more general two-user MISO broadcast correlated channel. The essential ingredients of the proposed scheme lie in the quantization and multicast of the overheard interferences, while broadcasting new private messages. Our proposed scheme smoothly bridges between the scheme recently proposed by Maddah-Ali and Tse with no current state information and a simple zero-forcing beamforming with perfect current state information.

Recent information theoretic results suggest that precoding on the multi-user downlink MIMO channel with delayed channel state information at the transmitter (CSIT) could lead to data rates much beyond the ones obtained without any CSIT, even in extreme situations when the delayed channel feedback is made totally obsolete by a feedback delay exceeding the channel coherence time. This surprising result is based on the ideas of interference repetition and alignment which allow the receivers to reconstruct information symbols which canceling out the interference completely, making it an optimal scheme in the infinite SNR regime. In this paper, we formulate a similar problem, yet at finite SNR. We propose a first construction for the precoder which matches the previous results at infinite SNR yet reaches a useful trade-off between interference alignment and signal enhancement at finite SNR, allowing for significant performance improvement in practical settings. We present two general precoding methods with arbitrary number of users by means of virtual MMSE and mutual information optimization, achieving good compromise between signal enhancement and interference alignment. Simulation results show substantial improvement due to the compromise between those two aspects.

Recent information theoretic results suggest that precoding on the multi-user downlink MIMO channel with delayed channel state information at the transmitter (CSIT) could lead to data rates much beyond the ones obtained without any CSIT, even in extreme situations when the delayed channel feedback is made totally obsolete by a feedback delay exceeding the channel coherence time. This surprising result is based on the ideas of interference forwarding and alignment which allow the receivers to reconstruct an information allowing them to cancel out the interference completely, making it an optimal scheme in the infinite SNR regime. In this paper, we formulate a similar problem, yet at finite SNR. We propose a new construction for the precoder which matches the previous results at infinite SNR yet reaches a useful tradeoff between interference alignment and signal enhancement at finite SNR, allowing for significant performance improvements in practical settings.

In practical downlink multiuser MIMO channels, users' channels are non-orthogonal, which results in inter-user interference and in turn degrades system performance. In this paper, a new subspace-domain linear transmit preprocessing technique is proposed in enhancing users' spatial separation and suppressing inter-user interference. We show by simulations that sum rate is significantly improved and its saturation is avoided.

We consider the time correlated MISO broadcast channel where the transmitter has partial knowledge on the current channel state, in addition to delayed channel state information (CSI). Rather than exploiting only the current CSI, as the zero-forcing precoding, or only the delayed CSI, as the Maddah-Ali-Tse (MAT) scheme, we propose a seamless strategy that takes advantage of both. The achievable degrees of freedom of the proposed scheme is characterized in terms of the quality of the current channel knowledge.

We consider the two-user MISO broadcast channel where the transmitter has imperfect knowledge on the current channel state, in addition to delayed channel state information. The degree of freedom region is completely characterized. The optimal scheme smoothly bridges between the scheme recently proposed by Maddah-Ali and Tse with no current state information and a simple zero-forcing beamforming with perfect current state information. The essential ingredients of this scheme lie in the quantization and multicasting of the overheard interferences, while broadcasting new private messages.