This paper investigates the joint estimation of direction-of-arrival (DOA) and polarization parameters using a one-dimensional polarization-sensitive mirrored array (1-D PS-MA) within the framework of parallel factor (PARAFAC) decomposition. The implicit rotational invariance property of the cosine-form spatial manifold is first derived, providing both a theoretical explanation for the formation of virtual signals during array expansion and an exact expression for their spatial manifold. Then, a full-rank hybrid co-array output is constructed via Toeplitz matrix reconstruction, and an explicit signal model is formulated based on the derived rotation invariance. To fully exploit polarization information, this model is incorporated into the PARAFAC decomposition framework. By analyzing the permutation and scaling ambiguities of the decomposed spatial and polarization manifolds, the proposed method enables efficient and automatically paired estimation of DOA and polarization parameters. Numerical results demonstrate that the proposed approach achieves higher estimation accuracy and resolution ability compared to several state-of-the-art methods.

Existing methods for solving grid mismatch problems are often based on first-order Taylor expansion, polynomial root-finding, or exhaustive search over a given angular range. However, these methods either approximate the original model inaccurately enough or are constrained by the search grid size, which leads to unsatisfactory DOA estimation accuracy. To address these issues, this paper proposes a high-accuracy off-grid DOA estimation algorithm based on the log-marginal likelihood function and the second-order Taylor expansion. The proposed algorithm derives the first- and second-order derivatives of the log-marginal likelihood function with respect to the target DOA and gets the numerical solution of the grid mismatch based on these derivatives. After obtaining the grid mismatch, the array manifold matrix as well as the signal and noise powers are updated iteratively, ultimately achieving off-grid DOA estimation. Simulation results demonstrate the superior DOA estimation performance of the proposed algorithm.

A two-stage cascaded direction of arrival (DOA) estimation framework is proposed for parameter estimation under non-ideal array conditions. The framework consists of two modules: contrastive reconstruction network (CRN) and DOA estimation network (DEN). To enhance robustness against array imperfections, CRN leverages self-supervised contrastive learning (SSCL) to recover covariance matrix, which is then fed into the DEN to perform the DOA estimation task. To further improve generalization, an augmentation strategy using non-ideal data, such as amplitude-phase perturbations and random sensor masking, is introduced during data training, where a two-stage training strategy is adopted to separately optimize the parameters of CRN and DEN. Simulation results demonstrate that the proposed framework achieves higher estimation accuracy and stronger robustness in non-ideal array conditions.

In this work, we consider the challenging gridless direction-of-arrival (DOA) estimation problem with arbitrary sparse arrays. By utilizing the feature relationship between the received array signals and the gridless sparse recovery method, a local-global feature extraction self-supervision network (LGFS-Net) is proposed to achieve gridless sparse recovery without labels. The proposed network consists of a local-global feature extraction module and a complex atomic norm minimization (ANM) self-supervision module. The feature extraction module models the relationship of array elements to recover the missing data corresponding to the holes of the virtual array by learning the local and global features between the received data of different array elements. The complex ANM self-supervision module is embedded in the ANM problem, which realizes gridless Toeplitz matrix recovery and ensures the complex data structure in each network layer by complex singular value decomposition (SVD), positive semidefinite constraints (PSD), and ANM loss. By training the network using the received data of an arbitrary sparse array and constructing a self-supervised loss function that does not require labels, the network exhibits excellent performance when there are holes in the virtual array, and its signal recovery and parameter estimation performance is similar to the case of uniform linear arrays and better than traditional gridless recovery methods, as demonstrated by computer simulations.

Direction-of-arrival (DOA) estimation using distributed arrays has emerged as a promising technique for autonomous vehicle (AV) positioning in Internet of Vehicles (IoV) systems. This article proposes a novel DOA estimation method based on orthogonal distributed arrays for accurate AV localization. Specifically, the covariance matrix between orthogonal arrays is exploited to construct two univariate polynomials for DOA estimation. As the polynomial degree depends on the array aperture and the discontinuous sensor layout, the resulting polynomials are often high-order and lacunary. To efficiently solve these polynomials, an improved differential-evolution (DE) algorithm is developed, featuring an adaptive mutation strategy to reduce computational cost and a counter-based mechanism to escape local optima. In addition, a covariance-based cost function is designed for 2-D angle pairing. The array aperture is further extended to enable simultaneous DOA estimation of multiple vehicles when the orthogonal arrays share a common sensor. Simulation results demonstrate that, compared with existing methods, the proposed approach achieves higher estimation accuracy and lower computational complexity, offering a promising solution for vehicle positioning in IoV environments.

This work addresses the challenge of received signal strength (RSS) based target localization in practical wireless sensor networks where non-line-of-sight (NLOS) propagation and sensor position uncertainty coexist. Unlike existing approaches that handle these impairments separately, a unified maximum likelihood formulation is developed to jointly model NLOS bias and sensor location errors. The resulting problem is reformulated and represented using a factor graph representation, with which an efficient iterative message passing algorithm is then developed to iteratively estimate the target location and sensor positions. Simulation results demonstrate that, compared to conventional convex relaxation-based approaches, the proposed one exhibits superior robustness and achieves the closest localization accuracy to the Cram´ er-Rao Lower Bound (CRLB) while reducing the computation time by three orders of magnitude.

In this paper, we investigate the question of which technology, fluid antenna systems (FAS) or active reconfigurable intelligent surfaces (ARIS), plays a more crucial role in FAS-ARIS wireless communication systems. To address this, we develop a comprehensive system model and explore the problem from an optimization perspective. We introduce an alternating optimization (AO) algorithm incorporating majorization-minimization (MM), successive convex approximation (SCA), and sequential rank-one constraint relaxation (SRCR) to tackle the non-convex challenges inherent in single-user scenario. Specifically, for the transmit beamforming of the BS optimization, we propose a closed-form rank-one solution with low-complexity. For the optimization the positions of fluid antennas (FAs) of the BS, the Taylor expansions and MM algorithm are utilized to construct the effective lower bounds and upper bounds of the objective function and constraints, transforming the non-convex optimization problem into a convex one. Furthermore, we use the SCA and SRCR to optimize the reflection coefficient matrix of the ARIS and effectively solve the rank-one constraint. To be more general, the proposed AO algorithm is then extended to multi-user scenario. Simulation results reveal that the relative importance of FAS and ARIS varies depending on the scenario: FAS proves more critical in simpler models with fewer reflecting elements or limited transmission paths, while ARIS becomes more significant in complex scenarios with a higher number of reflecting elements or transmission paths. Ultimately, the integration of both FAS and ARIS creates a win-win scenario, resulting in a more robust and efficient communication system. This study underscores the importance of combining FAS with ARIS, as their complementary use provides the most substantial benefits across different communication environments.

Large-scale multiple-input-multiple-output systems extend the Fresnel region, transforming the target source type from far-field to near-field (NF) with curved spherical wavefront, which poses challenges for accurate target localization. In this article, an exact NF polarization localization algorithm is proposed to estimate the 5-D parameters by establishing an L-shaped cocentered orthogonal loop and dipole (COLD) array. First, estimation of manifold matrices aligned with each coordinate axis is performed by forming the delay cross-correlation covariance matrices, and employing parallel factor decomposition. The unambiguous phases present in each submanifold matrix are employed to disambiguate the remaining phases, enabling the construction of linear equations in spatial geometry, from which linear least-squares parameter estimation is allowed. In particular, based on the rotational invariance relationship between the horizontal and vertical components received by the COLD array, the polarization parameters are estimated via total least squares. Under the exact NF polarization scenario, we derive the closed-form asymptotic parameter variances and the corresponding Cram & eacute;r-Rao bound as benchmark. The proposed algorithm operates without phase approximation and is free from quarter-wavelength sensor spacing limitation, and extensive simulations are provided to demonstrate its superiority over competing schemes in terms of parameter estimation performance.

This paper investigates an integrated sensing, communication, and computing network enabled by joint beamforming and cooperative device-to-device (D2D) and mobile edge computing (MEC) offloading. In the considered system, multiple full-duplex integrated sensing and communication devices perform target sensing while offloading partial latency-sensitive computation tasks through D2D and MEC links, thereby exploiting the complementary advantages of short-range D2D communication and the abundant computing resources of the MEC server. A D2D and MEC co-offloading framework is developed that simultaneously optimizes transceiver beamforming, task assignment, and computation resource allocation, with the aim of minimizing overall computation latency while ensuring sensing performance. To address the intractable issue of the formulated optimization problem, closed-form solutions of receive beamformers are first derived by solving minimum variance distortionless response problems. Subsequently, an alternating optimization algorithm is employed to decouple the original problem into two subproblems. To address their non-convexity, we apply semi-definite relaxation and successive convex approximation to transform each subproblem into a convex form and integrate penalty terms into the objective functions to promote rank-one solutions. Numerical results demonstrate that, compared to the state-of-the-art schemes, the proposed scheme achieves significant latency reduction under limited power and computing resources, while maintaining a high sensing signal-to-interference-plus-noise ratio.

In this paper, we address direction-of-arrival (DOA) estimation via non-coherent processing in partly calibrated arrays (PCAs) that consist of multiple subarrays with unknown orientation errors. In non-coherent processing, each subarray acts as a processing node that computes the local covariance matrix. However, the unknown orientation errors make these matrices no longer correspond to the same DOA. Although the traditional non-coherent MUSIC (Non-MUSIC) algorithm can solve this problem by directly increasing the search dimensions, the enormous computational complexity caused by high-dimensional search can not be avoided. Therefore, based on the geometric relationship between the reference subarray and other subarrays, utilizing the noise subspace of the reference subarray, we significantly reduce the impact of the orientation errors on the covariance matrices of other subarrays. Then, a new spectral function largely independent of the orientation errors is reconstructed, and the improved robustness non-coherent MUSIC algorithm (R-MUSIC) is proposed, where DOA estimation can be achieved without high-dimensional search. Thus, the computational complexity of the R-MUSIC is significantly lower than that of the directly applied Non-MUSIC. Furthermore, this geometric relationship can be expanded to the 2-dimensional (2-D) array for 2-D DOA estimation. Simulation results demonstrate the efficient and superior performance of R-MUSIC.

Multi-camera vehicle sensing provides a cost-effective solution for 3D environmental perception in intelligent vehicles. A practical challenge is that temporal fusion becomes unreliable when the ego-vehicle undergoes aggressive motion. Historical camera features are commonly aligned by rigid ego-pose warping, but residual motion-induced displacement can still degrade object localization, heading estimation, and velocity sensing. This paper presents GCL-BEV, a motion-aware temporal compensation framework for multi-camera vehicle sensor systems. The proposed framework uses synchronized surround-view cameras and ego-motion measurements as coupled sensing inputs. First, a Geometric-Aware Feature Enhancement (GAFE) module converts ego-motion priors into motion-conditioned BEV sampling offsets, allowing the visual sensing representation to compensate for local temporal misalignment before fusion. Second, a View-Consistency Learning (VCL) objective imposes a training-time equivariance constraint so that the sensor representation remains consistent under planar viewpoint perturbations. Across 10 random seeds on nuScenes, GCL-BEV achieves 57.80% ± 0.15 NDS and 46.22% ± 0.16 mAP with a ResNet-101 backbone. Compared with BEVDet4D, it reduces the mean Average Orientation Error by 5.4% and shows smaller degradation from steady driving to high-turn scenarios, indicating improved robustness for dynamic vehicle sensing.

Most existing direction-of-arrival (DOA) estimation methods are presented for a point source model, which could suffer from substantial performance degradation in multipath transmission scenarios. Although there have been some reports on DOA estimation in the case of multipath transmission or a distributed source model, they all rely on the prior knowledge of the number of sources under the assumption of uniform noise, which may not be available or satisfied in practice. This paper proposes two efficient two-dimensional (2-D) DOA estimation methods for incoherently distributed (ID) sources in unknown nonuniform noise, building on the generalized array manifold (GAM) of an L-shaped array and array covariance and cross covariance vectors. In particular, the first method employs a conjugate symmetry operation to enlarge the array aperture and two 1-D subspace spectral searches to estimate 2-D central DOAs independently. The second method constructs a sparse total least squares (STLS) problem to handle the impact of model bias and finite number of samples, and applies the alternating descent algorithm to achieve an improved 2-D central DOA estimation without knowing the number of sources. Finally, a simple and effective parameter pairing scheme is designed to avoid the ambiguity problem. Simulation results are provided to validate the effectiveness and superiority of the proposed methods.

Lane-level autonomous driving relies on high-accuracy vehicle positioning. Among different vehicle positioning approaches, direction-of-arrival (DOA)-based solutions are competitive as they avoid measuring delay information. However, a large distance between the vehicle and road side unit (RSU) compared with the antenna array aperture limits the positioning performance of the existing methods. To address this issue, in this letter, we propose an iterative positioning method using two collaborative RSUs and the spatial geometry, where the DOAs and the positions of vehicles are iteratively estimated. Numerical simulations demonstrate that the proposed method can achieve a positioning performance of millimeter-grade, improving the positioning performance by one to two orders of magnitude, compared with state-of-the-art methods.

This article addresses the problem of target localization in mixed line-of-sight (LOS) and non-LOS (NLOS) environments using differential time-delay (DTD) measurements from signals of opportunity emitted by a noncooperative transmitter with unknown position. The system operates without time synchronization among the transmitter and receivers, making it suitable for asynchronous multistatic localization. To mitigate the impact of outliers caused by NLOS propagation, the localization task is formulated as a least absolute deviation (LAD) optimization problem, and we then develop an efficient iterative message passing algorithm to solve it. Different from conventional methods, our proposed method circumvents the coupling issue between measurements and noises, making it suitable for dealing with outliers. Simulation results demonstrate that the proposed method outperforms state-of-the-art approaches under mixed LOS/NLOS conditions, exhibiting strong robustness and high accuracy even at high noise levels and with a limited number of sensors.

This article addresses the interference suppression problem in frequency-modulated continuous-wave radars. We propose an unsupervised learning framework based on an autoencoder architecture, which comprises two key modules: a unitary approximate message passing (UAMP)-based interference suppression module as the encoder and a vision transformer (ViT) reconstruction module as the decoder. For the interference suppression module, we propose a deep network by unrolling the UAMP for a sparse Bayesian learning algorithm to estimate the target signal, which exhibits sparsity in the frequency domain. The estimated target and interference components are subsequently fed into a ViT, which processes the input in patchwise fashion to capture global contextual dependencies for signal reconstruction, facilitating unsupervised learning without reliance on labeled data. To guide network training, we design a weighted loss function that combines mean-square error, sparsity control, and energy preservation, ensuring the retention of target power and sparsity while effectively suppressing interference. Both simulation and experimental results demonstrate the superior performance of the proposed network compared to state-of-the-art interference suppression approaches.

The dual-function radar and communication (DFRC) system addresses the problem of radio frequency congestion. Sparse arrays can reduce system cost and increase array aperture by constructing virtual coarrays. In this article, we consider the problem of receiving beamforming with sum and difference coarrays in the DFRC system. First, we present the radar and communication signal models with difference and sum coarrays in the DFRC system. Then, different receive beamforming methods are proposed to handle situations where radar targets and communication users are located at the same or different angles. When radar targets and communication users are at different angles, an optimal beamforming scheme is designed to assign them to separate beams. When a radar target and a communication user are at the same angle, we consider an orthogonality constraint to mitigate the radar and communication cross-interference in the same beam. We compared the beamforming performance of different sparse array structures, including coprime, nested array, nestedR array, and their virtual arrays, i.e., difference coarray or sum coarray. Simulation results demonstrate that the sparse array beamforming scheme enhances the resolution of receive beamforming in the DFRC system and reduces the cross-interference between received communication and radar signals.

This work addresses the maneuvering target localization problem with measurements of bistatic range (BR), bistatic range rate (BRR), and derivative of bistatic range rate (DBRR) from a distributed multiple-input multiple-output radar. To achieve high robustness to outliers, we adopt an l(1) norm-based approach and formulate a least absolute deviation (LAD) problem with heterogeneous measurements for jointly estimating target motion parameters of position, velocity, and acceleration. Then, we construct a factor graph representation by converting LAD into reweighted least squares problem and propose an iterative message passing localization algorithm. While demonstrated with BR, BRR, and DBRR, the proposed factor graph framework supports arbitrary heterogeneous measurement combinations. Cram & eacute;r-Rao Lower Bound is derived for performance evaluation. Simulations show that the proposed method has superior localization performance and robustness in the presence of measurement outliers and notably outperforms benchmark methods under moderate-to-high noise conditions. Results show that our proposed method delivers a performance improvement of over 25 dB compared to other methods, when the upper bound of BR outlier exceeds 50 m.

This paper investigates the three-dimensional (3D) localization of a moving target using hybrid angle-of-arrival (AOA) and time-difference-of-arrival (TDOA) measurements from a single fixed observer. Conventional pseudolinear estimators (PLE) suffer from significant biases, and existing bias-reduction methods, such as bias-reduced semidefinite programming (SDP) methods, often involve high computational complexity. Meanwhile, the Gauss-Newton (GN) method for solving the associated maximum likelihood estimation (MLE) problem is sensitive to initialization and may diverge under poor initial conditions or high measurement noise. To address these limitations, we propose batch and sequential regularized message passing (RMP) based iterative algorithms that have low-complexity and possess guaranteed convergence. We first formulate the problem as a regularized nonlinear least squares (NLS) optimization. Using eigenvalue decomposition (EVD) and factor graph (FG) representation, we propose a batch RMP algorithm in the empirical Bayesian framework. Then, we adopt a sequential regularized MLE estimation at each time step, and develop an FG-based sequential RMP algorithm to recursively update the target motion parameters. The proposed algorithms are rigorously analyzed in terms of convergence, mean square error (MSE), and bias. Simulation results validate the theoretical analyses and confirm the superior performance of the proposed RMP algorithms.

This paper proposes a secure and robust transceiver beamforming scheme to enhance the performance of device-to-device (D2D)-aided full-duplex integrated sensing and communication (ISAC) networks under imperfect channel state information (CSI). A joint optimization framework is developed to simultaneously optimize transceiver beamforming at the ISAC base station (BS) and transmit beamforming at D2D transmitters, with the aim of maximizing the worst-case sensing signal-to-interference-plus-noise ratio (SINR) while guaranteeing secure communication and quality-of-service (QoS) for cellular users (CUs) and D2D pairs. To address the intractable issue of the formulated joint optimization problem, we propose a robust transformation method based on the generalized S-lemma to convert CSI uncertainty constraints into tractable linear matrix inequalities (LMIs), enabling efficient handling of bounded channel errors. Subsequently, we propose an alternating optimization (AO) algorithm integrated with a double-checking strategy via semidefinite relaxation (SDR), where inner-layer feasibility verification and outer-layer rank-one validation ensure solution feasibility, and a rank penalty term accelerates convergence. Numerical results show that, compared to state-of-the-art schemes, the proposed scheme achieves significant performance improvements. These results confirm the robustness of the proposed method against complex interference and active eavesdropping threats, and highlight its superiority in balancing sensing-communication trade-offs for D2D-aided ISAC networks.

This work addresses the challenges of communication signal detection and direction of arrival (DOA) estimation in integrated sensing and communications (ISAC) systems with hardware imperfections. Conventional signal processing techniques often fail to effectively manage the complex nonlinearities caused by hardware imperfections, such as those introduced by power amplifiers and local oscillators. Recently, deep neural networks (DNNs) have been employed to mitigate the hardware imperfections, which however require a substantial amount of pilot signals for training, leading to unacceptable overhead and impracticality in fast time-varying channels. In this work, we employ an NN to characterize the nonlinear system, and propose a novel iterative approach to joint NN-based nonlinear system model learning, signal detection and DOA estimation. Instead of relying on pilot signals for NN learning, the proposed approach utilizes communication data signals as virtual training samples, enabling more accurate nonlinear model learning, which subsequently enhances signal detection and DOA estimation. A Bayesian framework is applied to the joint problem, wherein the NN parameters, the communication signals and the DOAs are jointly obtained by developing a message passing based inference algorithm. In particular, we impose sparse priors on the weights of the NN, so that overfitting can be better handled, resulting in significant improvement in system modeling performance. Extensive simulation results show that, compared to the state-of-the-art approaches, the proposed one delivers significantly better performance.