Sensing and communication are at the core of the Internet of Things, which usually function independently. For example, a smartphone can communicate over Wi-Fi or cellular networks while continuously acquiring sensory data from the environment through various sensors. This paper presents a novel framework, MuSAC (Mutualistic Sensing and Communication), which seamlessly integrates the collection of sensory data with existing communication systems, without adding any additional communication overhead (i.e., communication-free). The framework leverages the mutualistic relationship between specific communication data and sensory data to effectively crowdsource heterogeneous sensory data without harming communication performance in practical distributed systems. To embed massive sensory data into the current transmission of communication data, MuSAC presents novel neural networks to distill universal features from the raw data for compression at the sender side and then extract invariant features on the server side. By doing so, MuSAC eliminates additional communication overhead for sensory data collection while also mitigating privacy concerns and data heterogeneity in crowdsensing. To evaluate the performance of MuSAC, we first conduct system-level simulations in a distributed environment by embedding public human activity sensing datasets into cellular massive MIMO communication processes. The results demonstrate that MuSAC effectively supports heterogeneous sensory data collection over existing wireless communication links at zero additional communication overhead. We further implement a functional prototype of the MuSAC system using off-the-shelf devices, where real-world sensory data are seamlessly integrated into WiFi transmissions. This practical implementation demonstrates the viability of communication-free sensory data acquisition in real-world scenarios, paving the way for broader applications of MuSAC.

Integrated sensing and communication (ISAC) has emerged as a key enabler for next-generation wireless networks, supporting advanced applications such as high-precision localization and environment reconstruction. Cooperative ISAC (CoISAC) further enhances these capabilities by enabling multiple base stations (BSs) to jointly optimize communication and sensing performance through coordination. However, CoISAC beamforming design faces significant challenges due to system heterogeneity, large-scale problem complexity, and sensitivity to parameter estimation errors. Traditional deep learning-based techniques fail to exploit the unique structural characteristics of CoISAC systems, thereby limiting their ability to enhance system performance. To address these challenges, we propose a Link-Heterogeneous Graph Neural Network (LHGNN) for joint beamforming in CoISAC systems. Unlike conventional approaches, LHGNN models communication and sensing links as heterogeneous nodes and their interactions as edges, enabling the capture of the heterogeneous nature and intricate interactions of CoISAC systems. Furthermore, a graph attention mechanism is incorporated to dynamically adjust node and link importance, improving robustness to channel and position estimation errors. Numerical results demonstrate that the proposed attention-enhanced LHGNN achieves superior communication rates while maintaining sensing accuracy under power constraints. The proposed method also exhibits strong robustness to communication channel and position estimation error.

We focus on self-interference (SI) alleviated beamforming of 6 G full-duplex integrated sensing and communication (ISAC) systems, for achieving an on-demand sensing and communication performance tradeoff with suppressed SI for diverse user devices. However, SI-alleviated ISAC beamforming is of great challenge due to its complex problem structures and nonconvex optimization problem nature. In order to address this challenge, we propose to use the dual transformation framework for problem simplification, and exploit structured components of the problem model, such as convexity, linearity and fraction, for yielding an efficient iterative optimization solution. The proposed SI-alleviated beamforming method can gracefully take care of communication and sensing requirements with suppressed SI for diverse user devices, thus paving the way for an on-demand ISAC service. It is corroborated by numerical simulations that the proposed SI-alleviated beamforming method outperforms state-of-the-art ISAC beamforming baselines, due to our specially-tailored problem modeling and problem-specific algorithm design.

Deploying intelligent service and executing inference tasks in the proximity of the edge enable models to access enormous real-time data generated by the edge devices. However, the dilemma of fulfilling service demands with limited resources at edge devices impairs the efficacy of conventional data-oriented communication systems. To achieve a better trade-off between inference accuracy and communication overhead, in this paper, we propose a dynamic unmanned aerial vehicle (UAV)-assisted cooperative edge inference system, where a UAV acts as an edge server to aggregate the wide-view features from mobile sensors through Over-the-Air computation (AirComp) to complete the inference task cooperatively. Discriminant gain, an effective indicator for the inference accuracy, is adopted to realize task-oriented design. To exploit channel diversity and data diversity in the multi-device cooperative edge inference system, we maximize the discriminant gain of the AirComp feature aggregation by jointly optimizing the UAV trajectory and the power allocation policy with respect to the different important levels of feature dimensions. An alternating algorithm and a successive convex approximation (SCA)-based method are then proposed to solve the optimization problem. Numerical simulations further validate the efficacy of the proposed design compared to the baselines.

The utilization of reconfigurable intelligent meta-surface (RIS) for target localization has emerged as a prominent technology in sixth-generation (6G) wireless networks. Due to the deployment of large-scale RIS and the application of high-frequency signaling in 6G, future localization scenarios are anticipated to predominantly occur within the near-field region. In this paper, we investigate the scenario of near-field target localization supported by large-scale RIS with hybrid operational paradigms, wherein elements capable of passive or active signal processing (SP) are employed. Through a series of complex technical calculations, we derive the closed-form of Cramer-Rao bound (CRB) for near-field localization assisted by RIS with hybrid operational paradigms. By analyzing the Cramer-Rao bound (CRB) of near-field localization, we aim to optimize the ratio of active and passive SP elements of RIS to determine the optimal hybrid operational paradigms. To achieve this, we utilize a spatial angle approximation technique to convert the complex combinatorial optimization problem into a tractable optimization problem, which is amenable to efficient solutions using existing optimization algorithms. Through simulations, we show that the proposed technique can effectively find the optimal ratio of active and passive SP elements that minimizes the localization error.

Recently, many supervised deep learning-based methods have been investigated for large-scale MIMO channel estimation. However, the supervised methods require a large amount of ground truth channel as labeled data, which are difficult to obtain in practice. In this paper, we propose to learn the large-scale MIMO channel from compressed noisy measurements without any ground truth channel by introducing a loss function based on Generalized Stein’s Unbiased Risk Estimate (GSURE), which is an unbiased estimate of the projected mean-squared error (PMSE). Furthermore, we adopt deep equilibrium model (DEQ) to directly learn the equilibrium point of an iterative algorithm through an infinite-depth network, therefore it inherits the convergence performance of the underlying algorithm. We show that when using DEQ networks, employing GSURE for unsupervised learning can achieve performance close to that of supervised learning using MSE. Experiments demonstrate that the proposed GSURE-based unsupervised DEQ showcases superior channel estimation performance compared to various baselines when ground-truth channel is unavailable.

Sensing and communication are at the core of the Internet of Things, which usually function independently. For example, a smartphone can communicate over Wi-Fi or cellular networks while continuously acquiring sensory data from the environment through various sensors. This paper presents a novel framework, MuSAC (Mutualistic Sensing and Commu-nication), which seamlessly integrates the collection of sensory data with existing communication systems, without adding any extra communication overhead. The framework leverages the mutualistic relationship between specific communication data and sensory data to effectively crowdsource heterogeneous sensory data without harming communication performance in practical distributed systems. To embed massive sensory data into the current transmission of communication data, MuSAC presents novel neural networks to distill universal features from the raw data for compression at the sender side and then extract invariant features on the server side. By doing so, MuSAC eliminates additional communication costs for sensory data collection while also mitigating privacy concerns and data heterogeneity in crowd-sensing. Our real-world experimental validation in Wi-Fi and cellular Massive MIMO communication scenarios demonstrates the effectiveness of the MuSAC framework, shedding light on efficient mobile crowdsensing for massive IoT data collection.

Phase noise can cause significant performance degradation in multiple-input-multiple-output (MIMO) orthogonal frequency division multiplexing (OFDM) systems, especially for high-order data transmission. To mitigate the effect of phase noise on data transmission, pilot-based and blind-based algorithms are widely adopted in the existing works, which suffer from spectral efficiency degradation or formidable computational cost due to the large-scale and time-dependent properties of phase noise. In this paper, we propose an efficient multi-level maximum a posteriori (MMAP)-based blind data detection algorithm to address the phase noise compensation in MIMO-ODFM systems. The proposed algorithm, exploiting the spectral low-dimensional property of phase noise and the approximate message passing (AMP) rule, achieves a near optimal detection performance. The exploitation of low-pass characteristics of phase noise spectrum significantly reduces the computational complexity, and the adoption of AMP principle ensures a linear complexity of the algorithm with respect to the number of antennas and subcarriers. Thus, a good complexity-accuracy trade-off is obtained. Besides, the proposed algorithm is applicable to the scenarios of commonly shared oscillators and independent oscillators. The numerical experiments show that the proposed pilot-free data detection algorithm can achieve superior data transmission performance for channels with strong phase noise at a low complexity.

Channel estimation is crucial to modern wireless systems and becomes increasingly challenging when the ultra-sized antenna is configured in sub-6GHz wireless communication systems. In an ultra-massive multiple-input multiple-output (U-MIMO) orthogonal frequency division multiplex (OFDM) system, the channel demonstrates spatial non-stationarity. Additionally, the limited pilot location in the OFDM system further complicates the channel estimation process. In this paper, we propose a model-data dual-driven (MDD) scheme to jointly perform the model-driven non-stationary channel denoising and the data-driven channel interpolation in an end-to-end way, which is followed by a low-complexity channel refinement module to improve the robustness of the proposed scheme. Specifically, image contour extraction (ICE) is utilized to effectively eliminate the non-stationary noises in the channel matrices before being sent to the downstream interpolation network. An enhanced convolutional neural network (CNN)-based residual network (eCNN-RN) is developed to perform non-linear interpolations for recovering the U-MIMO-OFDM channels. Based on ICE, the proposed online refinement module can improve the generalizability of the learned model to a practical environment. Numerical experiments demonstrate the efficiency and the effectiveness of the cross-fertilization of the model-driven and data-driven approaches.

Channel state information (CSI) feedback has been a challenging task for downlink frequency division duplex (FDD) system. Meanwhile, sensory data collection in large-scale network is pivotal to support intelligent applications in the central server. In this paper, we propose a deep-learning-empowered adaptive CSI feedback compression and quantization based on the information-bottleneck principle, where the sensory data transmission is hidden within the CSI feedback to eliminate extra communication cost and preserve the data privacy at the same time. To reduce the impact of information hiding on CSI feedback, we focus on hiding data in the semantic level. The tradeoffs among communication efficiency, CSI accuracy, hidden information transfer accuracy, and privacy are jointly optimized. Simulations further verify that the proposed scheme can achieve accurate sensory data collection without resource occupation and accurate CSI feedback with limited feedback overhead simultaneously.

Conventional deep neural network (DNN)-based channel estimator is trained offline and deployed online with fixed weights, lacking the adaptivity and generality to fast varying channels. In this paper, we propose a real-time compressive channel tracking (CT) algorithm based on regularized deep generative model (DGM) to recover time-varying channels recursively. Specifically, we propose an untrained DGM for online CT, which does not require training over large datasets and can learn the channel variations on the fly. Moreover, DGM can capture the sparse structure of massive MIMO channels automatically. To further improve the tracking performance, we propose to regularize the DGM adaptively using a temporal prior to exploit the dynamic sparsity of channels, which is learned by an online trained sliding long short term with memory (LSTM) network. Simulations demonstrate the superior performance of the proposed algorithm compared to the model-based and supervised DNN-based CT algorithms.

In this paper, the positioning accuracy of multiple reconfigurable intelligent surface (RIS)-assisted multi-user direct localization system is studied. In particular, we incorporate statistical spatial correlation of multi-user locations and develop Bayesian Cramer-Rao bound (BCRB) to measure positioning accuracy. To address the information coupling between neighboring users induced by spatial correlation and position uncertainty, the equivalent Fisher information matrix (EFIM) of users' positions is decomposed to obtain a metric characterizing the efficiency of spatial correlation (EoSC). Random walk model (RWM) is adopted to provide a graph interpretation for EoSC, which unveils the information flow during positioning. Finally, we numerically illustrate the localization accuracy scaling law for several key system parameters.

Integrated sensing and communication (ISAC) has been a promising technology in beyond 5G and 6G network to simultaneously support high-speed information transfer and high-quality environmental perception. Cloud radio access networks (C-RAN) enables the cooperation among multiple base stations (BSs) to provide additional cooperation gains for both communication and sensing functionalities. In this paper, we investigate the cooperative ISAC (CoISAC) system with rate-splitting multiple access (RSMA) transmission scheme for advanced interference management and direct localization sensing scheme to provide high-accuracy positioning services. Specifically, we adopt Pareto optimization framework to characterize the achievable performance region of the CoISAC system built on the sum rate of multiple communication users and the positioning error bound of radar target. Then, we formulate communication-centric and radar-centric optimization problems in the CoISAC system to systematically search for the optimal Pareto boundary of the achievable performance region. Two iterative algorithms based on successive convex approximation (SCA) are proposed to effectively solve these two optimization problems, respectively, therefore near-optimal Pareto boundary can be found. Numerical results show that the RSMA-assisted CoISAC system achieves the best trade-off performance among various baselines. The cooperative scheme can fully unveil the potential of RSMA in the ISAC system by providing more freedom for rate-splitting policy design under limited resources. Therefore, incorporating RSMA in the CoISAC system can significantly boost the overall performance.

Channel sate acquisition in frequency-division du-plexing (FDD) massive MIMO system is challenging due to the huge feedback overhead. Machine learning (ML) has emerged as a powerful technology to address this challenge. In this paper, we resort to information bottleneck (IB) theoretical principle to design a joint feedback compression, quantization and channel learning algorithm in FDD massive MIMO systems, called IBNet. Compared to the existing ML-based designs, the proposed IBNet can systematically seek for the optimal balance between the channel estimation accuracy and feedback overhead. To auto-matically learn the feedback compression, a sparsity inducing prior is utilized to sparsify the feature vector, thereby reducing the feedback overhead significantly. Furthermore, to improve the generality of proposed IBNet, we propose an adaptive IBNet, which can adapt to different channel conditions with one neural network. Simulation results show that the proposed scheme significantly reduces the feedback overhead, meanwhile improving the channel estimation accuracy.

Accurate channel state information (CSI) acquisition with low pilot overhead has always been a problem for RIS-assisted massive MIMO systems. The design of phase shifts at RIS plays a key role in channel estimation (CE). In this paper, we incorporate the spatial feature of RIS into the design of a variational Bayesian inference-based CE algorithm, which is computational efficient and can exploit the sparse structure of cascaded channel. To optimally configure the phase shifts of RIS for CE, an optimal phase training algorithm and a simplified version are proposed by formulating the CE performance metric in terms of the RIS phases. The numerical simulations show that the proposed spatial feature aided optimal phase training of RIS as well as the compressive CE algorithm achieve superior CE performance with very low pilot overhead compared to the state-of-art baselines.

In this paper, we propose an untrained deep image prior-based compressive channel estimation (CE) algorithm in massive MIMO systems. We adopt deep convolutional generative network (DCGN) to learn the sparse prior of massive MIMO channels, whose weights can be optimized to match the compressed measurements. Unlike various supervised deep neural networks (DNN)-based CE algorithms, which require offline training over large simulated channel datasets, our DCGN does not require pre-training and can be learned online based on the real-time measurements to make it adaptable to the time-varying channels. Besides, DCGN can exploit the channel structure automatically without any prior knowledge. We further devise a learned regularization technique to improve the CE performance when the measurements are noisy and highly compressive. Simulations show that the proposed method can achieve more accurate and robust online CE performance than traditional compressive sensing (CS) and DNN-based methods.

Localization is an essential service for numerous applications. In particular, cooperative localization is a widely used technique for improving performance by utilizing information obtained from several base stations. In this paper, we study the potential of reconfigurable intelligent surfaces (RIS) for cooperative localization performance in mmWave MIMO systems. Firstly, we obtain the fundamental cooperative localization performance limit, i.e., Cramer-Rao lower bound (CRLB) based on the Fisher information. Then, we propose an optimal phase design at RIS to optimize the position accuracy. In particular, to handle the phase only constraints of RIS, an optimal passive beamforming (PBF) algorithm based on manifold optimization is proposed to minimize the CRLB, which, however, has a high complexity. Then, under mild conditions, we show that the CRLB minimization problem can be cast as a joint channel gain maximization problem, which enables a low-complexity closed-form PBF design at RIS. The simulation results show that the proposed optimal PBF for cooperative localization significantly improves the localization accuracy. Moreover, the proposed low-complexity PBF achieves near optimal performance with very low computational complexity.

This letter investigates a reconfigurable intelligent surface (RIS) assisted mmWave system for wireless area positioning. We aim to optimize the worst-case localization performance in terms of squared position error bound within the target region by jointly optimizing beamforming vectors at RIS and user equipment (UE), which leads to a nonconvex-nonconcave minimax problem. To tackle the challenging minimax problem, we propose a two-step algorithm with good convergence performance. Specifically, a joint array gain and path loss search (JAPS) algorithm is proposed to effectively find the exact solution of inner nonconcave maximization problem and a difference of convex (DC)-based algorithm is proposed to update the beamforming vectors at RIS and UE to improve the worst localization performance. Numerical results show that the proposed algorithm achieves significant performance gains over the benchmark schemes.

Compressive channel estimation (CCE) schemes have been proposed in millimeter-wave (mmWave) systems. The design of the analog combining matrix plays an important role to provide high channel recovery quality. In this paper, we propose a fixed-loading-based flexible hybrid beamforming (HBF) architecture (FL-FHA), which enables an arbitrary number of phase shifters (PSs) and a flexible connection strategy for the analog network. With the proposed flexible HBF architecture, we develop a variational-Bayesian-inference (VBI)-based algorithm to estimate the mmWave channel exploiting the channel support side information (CSSI) at the base station. Furthermore, we propose a parallel configuration optimization (CO) algorithm to optimize the FL-FHA according to CSSI, aiming at improving the overall CE performance. Simulation results show that the proposed CSSI-assisted joint CO and CCE outperforms the baselines with a small number of phase shifters.

In many applications in wireless communications, we need to recover a structured sparse signal from a linear measurement model with uncertain sensing matrix. There are two challenges of designing an algorithm framework for this problem. How to choose a flexible yet tractable sparse prior to capture different structured sparsities in specific applications? How to handle a sensing matrix with uncertain parameters and possibly correlated entries? As will be explained in the introduction, existing common methods in compressive sensing (CS), such as approximate message passing (AMP) and variational Bayesian inference (VBI), may not work well. To better address this problem, we propose a novel Turbo-VBI algorithm framework, in which a three-layer hierarchical structured (3LHS) sparse prior model is proposed to capture various structured sparsities that may occur in practice. By combining the message passing and VBI approaches via the turbo framework, the proposed Turbo-VBI algorithm is able to fully exploit the structured sparsity (as captured by the 3LHS sparse prior) for robust recovery of structured sparse signals under an uncertain sensing matrix. Finally, we apply the Turbo-VBI framework to solve two application problems in wireless communications and demonstrate its significant gain over the state-of-art CS algorithms.

Many important application scenarios in the future fifth generation (5G) systems, such as indoor navigation and autonomous driving, rely on accurate localization of users. In this paper, we propose a sparse-Bayesian-inference (SBI) based direct location algorithm for massive MIMO systems, which can exploit the sparse and high-resolution nature of angle of arrival (AoA) and any available statistical location information (SLI), to significantly improve the user localization accuracy. The existing common methods in SBI, such as approximate message passing (AMP) an variational Bayesian inference (VBI), may not work well for the massive MIMO localization problem due to their respective drawbacks. To overcome these drawbacks, we first propose a novel three-layer hierarchical structured (3LHS) sparse prior model to incorporate both the structured sparsity of the massive MIMO channel and the SLI into the SBI-based localization formulation. Then we propose a structured VBI algorithm called 3LHS-VBI to solve the resulting SBI-based localization problem. Finally, simulations verify the superior performance of the proposed location algorithm.

Channel estimation (CE) in massive MIMO system with limited RF chains is known to be a challenging problem. In many cases, base station (BS) can obtain certain channel support side information (CSSI), which can be exploited to enhance the CE performance and reduce the pilot overhead. Due to imperfectness of prior information, it's very important to utilize the CSSI in an optimal way to improve the CE performance. We propose an optimal-tuned weighted LASSO algorithm which can fully exploit the imperfect CSSI to optimize the CE performance in massive MIMO system with limited RF chains. In the proposed algorithm, the weighted l_{1} norm is used as the regularization function and the weights on the known support part (as indicated by the CSSI) and the rest are different. Based on the accuracy of CSSI, we obtain closed-form solution for the optimal LASSO weights which minimize the asymptotic normalized squared error (aNSE). Moreover, we derive closed-form expression of the minimum aNSE and characterize the minimum number of required pilots to achieve stable channel recovery. The theoretical analysis and simulation both show the performance advantages of our proposed solution over various baselines.

Recently, compressive channel estimation (CE) has been proposed to reduce the pilot overhead for massive MIMO with limited RF chains. One key issue is how to design the RF (analog) training vectors to achieve higher beamforming (BF) gain with fewer pilots. Specifically, narrow-beam RF training requires large pilot overhead for finding strongest paths, and random RF training suffers from low BF gain. We propose to use a mixture of narrow-beam and random RF training vectors, and exploit the channel support side information (CSSI) at the BS to do joint RF training and compressive CE. The narrow-beam RF training vectors are used to achieve a high BF gain, and the random RF training vectors are used to explore the unknown channel support to reduce the pilot overhead. Moreover, we derive closed-form bounds on the CE error. Both the analysis and simulations show that the proposed method can achieve substantial gains over various baseline methods.