Utilizing complex-valued neural networks (CVNNs) in wireless communication tasks has received growing attention for their ability to provide natural and effective representation of complex-valued signals and data. However, existing studies typically employ complex-valued versions of simple neural network architectures. Not only they merely scratch the surface of the extensive range of modern deep learning techniques, theoretical understanding of the superior performance of CVNNs is missing. To this end, this paper aims to fill both the theoretical and practice gap of employing CVNNs in wireless communications. In particular, we provide a comprehensive description on the various operations in CVNNs and theoretically prove that when the output dimension is large, the CVNN requires fewer layers than its real-valued counterpart to achieve a given approximation error of a continuous complex-valued function. Furthermore, to advance CVNNs in the field of wireless communications, this paper focuses on the transformer model, which represents a more sophisticated deep learning architecture and has been shown to have excellent performance in wireless communications but only in its real-valued form. In this aspect, we propose a fundamental paradigm of complex-valued transformers for wireless communications, including the complex-valued embedding module, encoding module, decoding module, and output projection module. Leveraging this structure, we develop customized complex-valued transformers for three representative applications in wireless communications: channel estimation, user activity detection, and joint design of pilot, feedback quantization, and precoder. These applications utilize transformers with varying levels of sophistication and span a variety of tasks, ranging from regression to classification, supervised to unsupervised learning, and specific module design to end-to-end design. Experimental results verify the theoretical advantage and effectiveness of the complex-valued transformers for the above three applications compared to other traditional real-valued neural network-based methods.

Joint activity detection and channel estimation is a crucial task in grant-free random access for massive machine-type communications. To enhance communication quality, intelligent reflecting surfaces (IRSs) have been proposed as a promising technology by controlling the propagation environment with passive reflecting elements. However, due to their passive nature and the non-Gaussian device-IRS-BS composite channels, IRSs introduce significant challenges for joint activity detection and channel estimation. To this end, this paper establishes an accurate statistical model for the composite channel, demonstrating that it follows a variance-gamma (VG) distribution. Based on the exact channel statistics, this paper employs a Bernoulli-VG prior and extends the standard approximate message passing algorithm to learn the Gamma-distributed channel variance within an expectation-maximization framework. Additionally, this paper introduces a novel approach to enforce consistency in device activity status across all base station antennas by transforming the activity detection task into the estimation of a dedicated active probability for each device. Extensive simulations validate the proposed VG channel model and demonstrate significant improvement due to imposing consistent activity probability across multiple antennas.

The recently emerged movable antenna (MA) shows great potential in leveraging spatial degrees of freedom for enhancing the performance of wireless systems. However, resource allocation in MA-aided systems faces unique challenges due to the non-convex and coupled constraints on antenna positions. This paper systematically reveals the challenges brought by the minimum MA separation constraints, and proposes a penalty framework for resource allocation under such new constraints in MA-aided systems. By introducing auxiliary variables, the proposed framework separates the non-convex and coupled antenna distance constraints from the movable region constraint. This enables the resulting problem be efficiently solved by alternating optimization, where the optimization of the original variables resembles that in conventional resource allocation problem while the optimization with respect to the auxiliary variables is achieved in closed-form solutions. To illustrate the effectiveness of the proposed framework, we present three case studies: capacity maximization, latency minimization, and regularized zero-forcing precoding. Simulation results demonstrate that the proposed optimization framework consistently outperforms state-of-the-art schemes.

A great amount of endeavor has recently been devoted to activity detection for massive machine-type communications in cell-free multiple-input multiple-output (MIMO) systems. However, as the number of antennas at the access points (APs) increases, the Rayleigh distance that separates the near-field and far-field regions also expands, rendering the conventional assumption of sole far-field propagation impractical. To address this challenge, this paper establishes a covariance-based formulation that can effectively capture the statistical property of hybrid near-far field channels. Based on this formulation, we theoretically reveal that increasing the proportion of near-field channels enhances the detection performance. Furthermore, we propose a distributed algorithm, where each AP performs local activity detection and only exchanges the detection results to the central processing unit, thus significantly reduces the computational complexity and the communication overhead. Not only with convergence guarantee, the proposed algorithm is unified in the sense that it can handle single-cell or cell-free systems with either near-field or far-field devices as special cases. Simulation results validate the theoretical analyses and demonstrate the superior performance of the proposed approach compared with existing methods.

This paper addresses the challenge of large model (LM)-embedded wireless network for handling the trade-off problem of model accuracy and network latency. To guarantee a high-quality of users’ service, the network latency should be minimized while maintaining an acceptable inference accuracy. To meet this requirement, LM quantization is proposed to reduce the latency. However, the excessive quantization may destroy the accuracy of LM inference. To this end, a promising fluid antenna (FA) technology is investigated for enhancing the transmission capacity, leading to a lower network latency in the LM-embedded multiple-input multiple-output (MIMO) network. To design the FA-assisted LM-embedded network with the lower latency and higher accuracy requirements, the latency and peak signal to-noise ratio (PSNR) are considered in the objective function. Then, an efficient optimization algorithm is proposed under the block coordinate descent framework. Simulation results are provided to show the convergence behavior of the proposed algorithm, and the performance gains from the proposed FA-assisted LM-embedded network over the other benchmark networks in terms of network latency and PSNR.

A great amount of endeavor has recently been devoted to activity detection for massive machine-type communications in cell-free massive MIMO. However, in practice, as the number of antennas at the access points (APs) increases, the Rayleigh distance that separates the near-field and far-field regions also expands, rendering the conventional assumption of far-field propagation alone impractical. To address this challenge, this paper considers a hybrid near-far field activity detection in cell-free massive MIMO, and establishes a covariance-based formulation, which facilitates the development of a distributed algorithm to alleviate the computational burden at the central processing unit (CPU). Specifically, each AP performs local activity detection for the devices and then transmits the detection result to the CPU for further processing. In particular, a novel coordinate descent algorithm based on the Sherman-Morrison-Woodbury update with Taylor expansion is proposed to handle the local detection problem at each AP. Moreover, we theoretically analyze how the hybrid near-far field channels affect the detection performance. Simulation results validate the theoretical analysis and demonstrate the superior performance of the proposed approach compared with existing approaches.

To improve the generalization of the autonomous driving (AD) perception model, vehicles need to update the model over time based on the continuously collected data. As time progresses, the amount of data fitted by the AD model expands, which helps to improve the AD model generalization substantially. However, such ever-expanding data is a double-edged sword for the AD model. Specifically, as the fitted data volume grows to exceed the AD model’s fitting capacities, the AD model is prone to under-fitting. To address this issue, we propose to use a pretrained Large Vision Models (LVMs) as backbone coupled with downstream perception head to understand AD semantic information. This design can not only surmount the aforementioned under-fitting problem due to LVMs’ powerful fitting capabilities, but also enhance the perception generalization thanks to LVMs’ vast and diverse training data. On the other hand, to mitigate vehicles’ computational burden of training the perception head while running LVM backbone, we introduce a Posterior Optimization Trajectory (POT)-Guided optimization scheme (POTGui) to accelerate the convergence. Concretely, we propose a POT Generator (POTGen) to generate posterior (future) optimization direction in advance to guide the current optimization iteration, through which the model can generally converge within 10 epochs. Extensive experiments demonstrate that the proposed method improves the performance by over 66.48% and converges faster over 6 times, compared to the existing state-of-the-art approaches.

Vehicle driving simulators have been widely used in fields including road design, automotive development, and driver training. As a core component of the simulators, motion cueing algorithms (MCAs) aim to reproduce realistic vehicle motion sensations while respecting workspace limitations for a high-fidelity driving experience. Recent advancements have identified model predictive control (MPC) as a promising approach for MCA. However, conventional MPC-based MCA faces computational bottlenecks that limit its performance. Although recent studies have employed swarm intelligence optimization algorithms, such as the genetic algorithm and grey wolf optimizer, to enhance the performance of MPC-based MCA through horizon adjustments, the improvement path from the computational cost perspective is still overlooked. Therefore, this paper proposes a costs-optimization framework for MPC-based MCA (COMPC-based MCA), which integrates two key components: an Operator Splitting Quadratic Program (OSQP) solver and a Snow Ablation Optimizer (SAO) that considers computational costs and motion sensation errors for parameter selection. Experiment results validated the performance of the proposed framework. COMPC-based MCA can achieve fast quadratic programming (QP) solving and select appropriate parameters for performance improvement while considering computational costs.

Deep learning-based Autonomous Driving (AD) perception models often exhibit poor generalization due to data heterogeneity in an ever domain-shifting environment. While Federated Learning (FL) could improve the generalization of an AD model (known as FedAD system), conventional models often struggle with under-fitting as the amount of accumulated training data progressively increases. To address this issue, instead of conventional small models, employing Large Vision Models (LVMs) in FedAD is a viable option for better learning of representations from a vast volume of data. However, implementing LVMs in FedAD introduces three challenges:(I)the extremely high communication overheads associated with transmitting LVMs between participating vehicles and a central server;(II)lack of computing resource to deploy LVMs on each vehicle;(III)the performance drop due to LVM focusing on shared features but overlooking local vehicle characteristics. To overcome these challenges, we propose pFedLVM, a LVM-Driven, Latent Feature-Based Personalized Federated Learning framework. In this approach, the LVM is deployed only on central server, which effectively alleviates the computational burden on individual vehicles. Furthermore, the exchange between central server and vehicles are the learned features rather than the LVM parameters, which significantly reduces communication overhead. In addition, we utilize both shared features from all participating vehicles and individual characteristics from each vehicle to establish a personalized learning mechanism. This enables each vehicle’s model to learn features from others while preserving its personalized characteristics, thereby outperforming globally shared models trained in general FL. As a demonstration of the proposed pFedLVM, this paper focuses on the semantic segmentation (SSeg) task. Extensive experiments demonstrate that pFedLVM outperforms the existing state-of-the-art approach by 18.47%, 25.60%, 51.03% and 14.19% in terms of mIoU, mF1, mPrecision and mRecall, respectively.

Street Scene Semantic Understanding (denoted as TriSU) is a complex task for autonomous driving (AD). However, inference model trained from data in a particular geographical region faces poor generalization when applied in other regions due to inter-city data domain-shift. Hierarchical Federated Learning (HFL) offers a potential solution for improving TriSU model generalization by collaborative privacy-preserving training over distributed datasets from different cities. Unfortunately, it suffers from slow convergence because the data from different cities are with disparate statistical properties. Going beyond existing HFL methods, we propose a Gaussian heterogeneous HFL algorithm (FedGau) to address inter-city data heterogeneity so that convergence can be accelerated. In the proposed FedGau algorithm, both single RGB image and RGB dataset are modelled as Gaussian distributions for aggregation weight design. This approach not only differentiates each RGB image by respective statistical distribution, but also exploits the statistics of dataset from each city in addition to the conventionally considered data volume. With the proposed approach, the convergence is accelerated by 35.5%-40.6% compared to existing state-of-the-art (SOTA) HFL methods. On the other hand, to reduce the involved communication resource, we further introduce a novel performance-aware adaptive resource scheduling (AdapRS) policy. Unlike the traditional static resource scheduling policy that exchanges a fixed number of models between two adjacent aggregations, AdapRS adjusts the number of model aggregation at different levels of HFL so that unnecessary communications are minimized. Extensive experiments demonstrate that AdapRS saves 29.65% communication overhead compared to conventional static resource scheduling policy while maintaining almost the same performance.

The key technologies of sixth generation (6G), such as ultra-massive multiple-input multiple-output (MIMO), enable intricate interactions between antennas and wireless propagation environments. As a result, it becomes necessary to develop joint models that encompass both antennas and wireless propagation channels. To achieve this, we utilize the multi-port communication theory, which considers impedance matching among the source, transmission medium, and load to facilitate efficient power transfer. Specifically, we first investigate the impact of insertion loss, mutual coupling, and other factors on the performance of multi-port matching networks. Next, to further improve system performance, we explore two important deep unfolding designs for the multi-port matching networks: beamforming and power control, respectively. For the hybrid beamforming, we develop a deep unfolding framework, i.e., projected gradient descent (PGD)-Net based on unfolding projected gradient descent. For the power control, we design a deep unfolding network, graph neural network (GNN) aided alternating optimization (AO)-Net, which considers the interaction between different ports in optimizing power allocation. Numerical results verify the necessity of considering insertion loss in the dynamic metasurface antenna (DMA) performance analysis. Besides, the proposed PGD-Net based hybrid beamforming approaches approximate the conventional model-based algorithm with very low complexity. Moreover, our proposed power control scheme has a fast run time compared to the traditional weighted minimum mean squared error (WMMSE) method.

Activity detection is an important task in the next generation Internet-of-things (IoT) networks. Existing algorithms mostly require precise information about the network, such as large-scale fading, noise variance, and small-scale fading statistics. Acquiring such information would take a significant overhead and their estimated values might not be accurate. This problem is even more severe in cell-free networks as more parameters are acquired. Therefore, this paper sets out to investigate this problem without the above mentioned information. In order to handle so many unknown parameters, this paper employs a Bayesian approach, where they are endowed with prior distributions as regularizations. Together with the likelihood function, a maximum a posteriori (MAP) estimator is derived. Simulations demonstrate that the proposed method outperforms state-of-the-art methods especially under imprecise information.

Street Scene Semantic Understanding (denoted as TriSU) is a crucial but complex task for world-wide distributed autonomous driving (AD) vehicles (e.g., Tesla). Its inference model faces poor generalization issue due to inter-city domain-shift. Hierarchical Federated Learning (HFL) offers a potential solution for improving TriSU model generalization, but suffers from slow convergence rate because of vehicles’ surrounding heterogeneity across cities. Going beyond existing HFL works that have deficient capabilities in complex tasks, we propose a rapid-converged heterogeneous HFL framework (FedRC) to address the inter-city data heterogeneity and accelerate HFL model convergence rate. In our proposed FedRC framework, both single RGB image and RGB dataset are modelled as Gaussian distributions in HFL aggregation weight design. This approach not only differentiates each RGB sample instead of typically equalizing them, but also considers both data volume and statistical properties rather than simply taking data quantity into consideration. Extensive experiments on the TriSU task using across-city datasets demonstrate that FedRC converges faster than the state-of-the-art benchmark by 38.7%, 37.5%, 35.5%, and 40.6% in terms of mIoU, mPrecision, mRecall, and mF1, respectively. Furthermore, qualitative evaluations in the CARLA simulation environment confirm that the proposed FedRC framework delivers top-tier performance.

Reconfigurable intelligent surface (RIS) has the capability to significantly enhance physical layer security by reconfiguring the propagation in wireless communications. However, due to the cascaded channel brought by the RIS and the hostile nature of potential eavesdroppers, acquiring perfect channel state information (CSI) of the eavesdroppers is challenging. Worse still, if the eavesdroppers are equipped with multiple antennas and there exists spatial correlation at the RIS due to closely spaced RIS elements, the random channel matrices are complicatedly coupled with the phase shift and other wireless resources in the outage probabilistic constraint, making their optimizations intractable. To date, there has been no systematic and feasible approach to address such a challenge. To fill this gap, this paper for the first time reveals an analytical transformation for handling the intractable outage probabilistic constraint. It is theoretically established that when the maximum tolerable outage probability is smaller than a threshold around 0.4, which generally holds in practice, the proposed transformation is exact and suffers no performance loss. As an illustrative example of the developed constraint transformation, the secure energy efficiency maximization is selected as the objecitve function and the resultant resource optimization is handled by the alternating maximization framework. Numerical results are presented to show the rapid convergence behavior of the proposed algorithm and unveil that the proposed probabilistic constraint transformation has superiority over the Bernstein-Type Inequality approximation. Compared with several baseline schemes (e.g., random phase-shift, fixed phase-shift, RIS ignoring CSI uncertainty, and secure transmission without RIS), the proposed scheme significantly boosts the performance, underscoring the significance of appropriately managing the probabilistic constraint outage and optimizing RIS phase shifts for secure transmission against multi-antenna eavesdroppers.

A great amount of endeavour has recently been devoted to activity detection for cell-free massive multiple-input multiple-output (MIMO) systems, where multiple access points (APs) jointly identify the active devices from a large number of potential devices. In practice, the APs and the central processing unit (CPU) are connected by capacity-limited fronthauls and the signals at the APs need to be compressed/quantized before they are forwarded to the CPU. However, existing approaches treat the compression/quantization and activity detection as separate tasks, which makes it difficult to achieve global system optimality. To tackle the above problem, this paper proposes an augmented model-driven end-to-end learning framework which jointly optimizes the compression modules, quantization modules at the APs, and the decompression module and detection module at the CPU. Specifically, deep unfolding is leveraged for designing the detection module in order to inherit the domain knowledge derived from the optimization algorithm, and other modules are constructed by judiciously designed neural network architectures for improving the learning capability. Furthermore, we design an enhanced scheme so that the proposed framework is adaptable to different compression rates. We demonstrate numerically that the proposed framework significantly reduces the computational complexity and achieves better detection performance than the conventional approaches. Moreover, it costs a much smaller number of bits on the fronthauls while still maintaining the detection performance.

This paper investigates a covariance-based approach for intelligent reflecting surface (IRS) aided activity detection in massive machine-type communications (mMTC). In the conventional scenario without IRS, the covariance-based approach, which exploits the probability density function (PDF) of the received signals at the base station (BS), has been demonstrated to outperform the compressed sensing approach. However, when taking the impact of the IRS into account, due to the newly introduced cascaded channels, it is difficult to obtain the exact PDF of the received signals at the BS. To tackle this challenge, we propose an approximation for the intended PDF with tunable parameters in the covariance matrix of the received signals. Based on the proposed tractable reformulation, an analytic framework is established to reveal the guideline for the phase shift design. Moreover, to determine the optimal correlation parameters, a deep unfolding approach is further leveraged by regarding them as trainable parameters. Simulation results validate the theoretical analysis and demonstrate the superior performance of the proposed learning approach.

Reconfigurable intelligent surface (RIS) has the potential to significantly enhance the physical layer security by reconfiguring the wireless propagation environment. However, due to the hostile nature of potential eavesdroppers and the cascaded channel brought by the RIS, acquiring perfect channel state information (CSI) of the eavesdroppers is challenging. Worse still, if the eavesdroppers are equipped with multiple antennas, the design of the optimal phase-shift, power allocation, and secure transmission data rate are intractable due to the couplings of random channel matrices in the outage probability. To overcome these challenges, this paper for the first time reveals an analytical transformation for handling the outage probabilistic constraint in the secure energy efficiency maximization problem due to multi-antenna eavesdropper. The resultant problem is readily handled under the alternating maximization framework. Simulation results unveil that the proposed probabilistic constraint transformation and the associated optimization algorithm provide superior secure energy efficiency over the baseline schemes of random phase-shift, fixed phase-shift, RIS ignoring CSI uncertainty, and secure transmission without RIS.

This paper investigates a covariance-based learning approach for intelligent reflecting surface (IRS) aided activity detection in massive machine-type communications (mMTC). In the conventional scenario without IRS, the covariance-based approach has been demonstrated to outperform the compressed sensing approach, as the covariance-based approach can well exploit the probability density function (PDF) of the received signals at the base station (BS). However, when taking the impact of the IRS into account, due to the newly introduced cascaded channels, it is quite difficult to obtain the exact PDF of the received signals at the BS. To tackle this challenge, we propose an approximation for the intended PDF by modeling a correlation parameter in the covariance matrix of the received signals. Based on the covariance-based formulation, a learning approach is further proposed to automatically learn the correlation parameter. Simulation results demonstrate the performance of the covariance-based activity detection, and the superiority of the proposed covariance-based learning approach.

Device activity detection in the emerging cell-free massive multiple-input multiple-output systems has been recognized as a crucial task in machine-type communications, in which multiple access points jointly identify the active devices from a large number of potential devices based on the received signals. Most of the existing works addressing this problem rely on the impractical assumption that different active devices transmit signals synchronously. However, in practice, synchronization cannot be guaranteed due to the low-cost oscillators, which brings additional discontinuous and nonconvex constraints to the detection problem. To address this challenge, this paper reveals an equivalent reformulation to the asynchronous activity detection problem, which facilitates the development of a distributed algorithm that satisfies the highly nonconvex constraints in a gentle fashion as the iteration number increases. To reduce the capacity requirements of the fronthauls, we further design a communication-efficient accelerated distributed algorithm. Simulation results demonstrate that the proposed two distributed algorithms outperform state-of-the-art approaches. Moreover, the accelerated distributed algorithm requires a very small number of quantization bits to approach the ideal detection performance.

A great amount of endeavour has recently been devoted to device activity detection in massive machine-type communications. This paper targets at a practical issue: communication-efficient joint signal compression and activity detection in cell-free massive MIMO with capacity-limited fronthauls. To this end, we propose a novel deep learning framework which jointly optimizes the compression modules, quantization modules at the access points, and the decompression module and detection module at the central processing unit. Specifically, deep unfolding is leveraged for designing the detection module in order to inherit the domain knowledge derived from the optimization algorithm, and the other modules are constructed by generic layers for increasing the learning capability. A joint training strategy is proposed to optimize all the modules in an end-to-end manner. Numerical results demonstrate the superiority of the proposed end-to-end learning framework compared with classical optimization methods.

Device activity detection in the emerging cell-free massive multiple-input multiple-output (MIMO) systems has been recognized as a crucial task in machine-type communications, in which multiple access points (APs) jointly identify the active devices from a large number of potential devices based on the received signals. Most of the existing works addressing this problem rely on the impractical assumption that different active devices transmit signals synchronously. However, in practice, synchronization cannot be guaranteed due to the low-cost oscillators, which brings additional discontinuous and nonconvex constraints to the detection problem. To address this challenge, this paper reveals an equivalent reformulation to the asynchronous activity detection problem, which facilitates the development of a centralized algorithm and a distributed algorithm that satisfy the highly nonconvex constraints in a gentle fashion as the iteration number increases, so that the sequence generated by the proposed algorithms can get around bad stationary points. To reduce the capacity requirements of the fronthauls, we further design a communication-efficient accelerated distributed algorithm. Simulation results demonstrate that the proposed centralized and distributed algorithms outperform state-of-the-art approaches, and the proposed accelerated distributed algorithm achieves close detection performance to that of the centralized algorithm but with a much smaller number of bits to be transmitted on the fronthaul links.

Grant-free random access is a promising mechanism to support modern massive machine-type communications in which devices are sporadically active with small payloads. Under this random access, a unique challenge is the detection of device activity without the cooperation from devices. Furthermore, for only a few bits of data, it is more efficient to embed the data to the signature sequences so that the activity and data detection can be jointly carried out. However, compared with the vanilla device activity detection, joint activity and data detection has an extra discontinuous sparsity constraint, which makes the detection problem more challenging. In contrast to the prevalent way of first neglecting the discontinuous sparsity constraint and reinforcing it at the end, this paper proposes a novel approach to incorporate the discontinuous sparsity constraint into the optimization procedure. In particular, we first establish the equivalence between the discontinuous sparsity constraint and a continuous difference-of-norms (DN) form. Then, by introducing a DN penalty term in the objective function, an iterative DN penalty method with an increasing penalty weight is adopted. We prove theoretically that by solving each penalized problem to a stationary solution, the discontinuous sparsity constraint can be exactly satisfied when the penalty weight is sufficiently large, and the resulting solution is guaranteed to be at least a stationary point of the original problem. Due to the superior theoretical guarantee, simulation results demonstrate that the proposed method achieves around 10 times better detection performance than state-of-the-art approaches.