Code division multiple access (CDMA) is missing in 4G and 5G systems, albeit it possesses many salient technical advantages over OFDMA, such as its unique multipath diversity gain with RAKE receiver. This work aims to propose a way to vitalize the performance of CDMA systems with M-ary code modulation (i.e., M-ary CDMA). In particular, non-orthogonal multiple access (NOMA) technique is leveraged to augment spectrum efficiency. Built upon the foundation of a high spectrum efficiency of M-ary CDMA systems, NOMA is integrated with M-ary CDMA for enhancing its user capacity further. Specifically, this work showcases the designs of downlink transmitter and receiver structures for a NOMA based M-ary CDMA system, and validates the effectiveness of the proposed scheme in terms of improved user capacity and bit error rate via simulations. Finally, the challenges and limitations faced by the proposed scheme are identified and the directions for future research are given.

As the Internet of Things (IoT) systems continue to evolve toward higher levels of intelligence, autonomy, and adaptability, the boundary between communication, sensing, computation, and control is becoming increasingly blurred. This transition requires us to rethink some fundamental issues, ranging from ethical design, energy sustainability, Artificial Intelligence, spectrum efficiency, etc. The IEEE Communications Magazine publishes the Internet of Things (IoT) Series to capture these transformative shifts, providing a platform for cutting-edge research that addresses the critical challenges of ethics, sustainability, efficiency, and integration. In this issue, we are pleased to present four rigorously reviewed articles that span the spectrum of next-generation IoT development. From ethical Radio Frequency (RF) holography and intelligent wireless charging to 6G-enabled Artificial Intelligence of Things (AIoT) architecture and antenna polarization technique, these contributions collectively pave the roadmap toward a more efficient wireless future.

This paper investigates a fairness optimization problem of physical layer security (PLS) in an intelligent reflecting surface (IRS) assisted multi-user uplink communication system. Due to the lack of eavesdroppers' instantaneous channel state information (CSI), we adopt secrecy outage probability (SOP) as a security performance metric, for which a min-max SOP minimization problem is formulated. To solve this problem efficiently, an alternating optimization (AO) scheme is proposed, which combines smoothing techniques with a fast iterative shrinkage thresholding algorithm (FISTA) to alternately optimize receive beamforming vectors and IRS phase shift matrix. Compared to conventional semidefinite relaxation (SDR) method, the proposed scheme reduces computational overhead significantly. Simulation results confirm that it not only achieves a significantly lower SOP but also improves computational efficiency substantially.

Network softwarization, enabled by Software- Defined Networking (SDN), Network Function Virtualization (NFV), and cloud-native architectures, accelerates service innovation but also amplifies operational complexity in multi-domain and multi-vendor environments. Recent advances in Large Language Models (LLMs) offer new opportunities for Artificial Intelligence IT Operations (AIOps) and intent-based network management, but their practical use remains limited by hallucinations and difficulties of grounding decisions in rapidly evolving operational knowledge, including policies, configurations, runbooks, trouble tickets, and telemetry. This article presents an evidence-grounded dynamic hierarchical Retrieval-Augmented Generation (RAG) reference architecture for intent-driven network softwarization and management. The architecture routes operator requests by intent, performs progressive evidence retrieval from source selection to snippet grounding, and invokes a feedback-driven adequacy loop when evidence is insufficient, fragmented, stale, or inconsistent. For action-sensitive requests, it adds execution guardrails that require evidence adequacy, affected-scope identification, rollback or verification planning, and explicit human-machine interaction confirmation before any high-impact write-back or controller-facing action is allowed. Rather than claiming exhaustive telecom-scale validation, this article contributes a deployment-oriented blueprint supported by compact evidence. The current prototype achieved 87.5% lookup-versus-diagnostic routing accuracy, and a scenario-grounded mini-corpus validation illustrates how all five profiles guide evidence ranking, response composition, adequacy checking, and guardrail activation. The proposed framework offers a practical path toward more trustworthy AI-assisted management in real-world softwarized networks.

In Internet of Things (IoT) edge inference systems, Softmax is a core component in multiclass classification heads and attention mechanisms. In resource-constrained IoT edge devices, conventional numerically stable implementations require multiple passes over the data, which increases memory bandwidth pressure and leads to hardly predictable variable processing latency. This article presents conditional online Softmax, a control-flow-predictable algorithm that fuses reduction and normalization into a single pass to cut memory traffic while preserving numerical stability. We pair it with a branch-free vectorized kernel implemented using RISC-V vector (RVV) extension intrinsics, which better exploits instruction-level parallelism on low-core-count edge processors. On a standard RISC-V edge platform, conditional online Softmax achieves a 7% speedup over the numerically stable safe Softmax and a 43% speedup over the original online Softmax. Furthermore, manual RVV vectorization delivers a 2.9 & times; speedup compared to GNU Compiler Collection (GCC) auto-vectorization. The resulting design reduces cache misses and improves latency predictability, making it a drop-in replacement for Softmax in streaming and microbatch pipelines commonly found in IoT gateways and sensor nodes. The implementation is portable across vector length (VLEN) and standard element width (SEW) settings and requires no specialized accelerators, which simplifies its deployment in practical IoT edge inference systems.

Spectrum resources are indispensable for the deployment of wireless communications and they are currently facing significant scarcity issues. Dynamic spectrum access (DSA) plays an important role in addressing this challenge. Machine learning approaches can be used in the development of spectrum access technologies in cognitive radios. However, heterogeneity of secondary users (SUs) has been overlooked in most previous studies, particularly in cognitive radio networks, where SUs possess varying observation spaces, transmit powers, and aggregation capabilities for idle channels. In this work, we establish a dynamic spectrum access system model with inhomogeneous SUs and propose a novel solution for multi-user dynamic spectrum access that integrates a heterogeneous agent deep reinforcement learning algorithm to address inhomogeneous challenges. Simulation results demonstrate that the heterogeneous-agent reinforcement learning (HARL) based dynamic spectrum access scheme outperforms other prevalent multi-agent deep reinforcement learning approaches. The proposed scheme effectively enhances network throughput, reduces collisions, and addresses the user-inhomogeneity issues.

Integrated sensing and communication (ISAC) has emerged as a critical convergence of traditional communication and radar systems in terms of spectrum, software, and hardware platforms within a unified framework. This survey aims to provide a comprehensive overview on ISAC, beginning with its layered architecture and continuing to discuss about many key components, such as technologies, standardizations, prototypes and testbeds. ISAC architecture is divided into terminal and signal layer, resource layer, function layer, application layer, and computation layer, each of which is examined in detail to understand its roles and interplay within an ISAC ecosystem. Moving beyond the layered architecture, we explore various enabling technologies that empower ISAC, such as beamforming, antenna array design, etc. Notably, this paper reviews emerging technologies such as reconfigurable intelligent surfaces (RIS), new materials, and innovative approach of software-defined radio/radar (SDR2), as well as near-field (NF) ISAC. In addition, this survey covers the standardization efforts of international bodies, which are instrumental in shaping an evolutional trajectory of ISAC technologies. The prototypes and testbeds are presented to validate theoretical concepts and practical implementations of ISAC. Finally, the survey concludes with open issues and potential future directions for ISAC, providing a roadmap for its ongoing development and integration into next-generation networks.

Reinforcement learning (RL) is a promising AI algorithm supporting latency-sensitive applications in 6G to enable ultra-reliable low-latency communication (URLLC). This work focuses on RL-assisted real-time resource allocation in mobile communications. Location and mobility of cellular user equipments (CUEs) are critical information for rapid convergence of resource allocation, in which deep Q-network (DQN) algorithm is useful to facilitate resource allocation to maximize sum rate of a cellular network, whose up-link channels are shared by device-to-device (D2D) UEs (DUEs). To ensure signal-to-interference-plus-noise ratios (SINRs) of both CUEs and DUEs, CUE’s guaranteed and prohibited areas are defined to align CUE locations with DUEs’ transmit powers for precise power allocation. The simulation results show that the proposed RL approach offers a fast tracking convergence with a balanced quality of services (QoSs) for both CUEs and DUEs. The application of the DQN-based algorithm can be extended beyond D2D communications to support low-latency location/mobility-aware communications.

It is a challenge to accurately model the propagation with multilayer structures, such as stacked intelligent metasurfaces (SIMs) in electromagnetics research, as conventional models fail to capture dominant diffraction and interlayer coupling effects. To address this challenge, this article develops a physics-based propagation framework grounded in vector Rayleigh-Sommerfeld diffraction integrals. The framework is complemented by an equivalent spatial spectrum formulation based on plane-wave decomposition, which reframes diffraction as a spatial filtering process within linear system theory. For practical applicability, it also incorporates computationally efficient near- and far-field approximations governed by an adaptive phase-error-controlled boundary. To demonstrate its engineering relevance, the framework is applied in electromagnetic information theory (EIT). In an SIM-MIMO system, it establishes a direct link between the diffraction Green's function and the channel matrix. This connection enables the derivation of mutual information (MI) and reveals fundamental tradeoffs between physical parameters and channel capacity. In summary, the proposed approach not only establishes a rigorous link between electromagnetic modeling and system performance but also provides physics-based co-design guidelines that challenge conventional assumptions and enable the optimization of next-generation SIM-MIMO architectures.

Cell-free massive multiple-input multiple-output (CFmMIMO) is one of the key enabling technologies for massive machine type communication (mMTC), where user activity detection and channel estimation face a significant challenge due to asynchronous transmissions caused by either propagation delays among access points (APs) or low-cost oscillators. This work aims to propose a joint user activity detection and channel estimate approach for asynchronous mMTC scenarios in CFmMIMO. We deal with user activity detection and channel estimation as a structural compressive sensing problem, which is related to asynchronous latency, active users, and channel coefficients, making use of sporadic nature of mMTC traffic. In particular, we implement an asynchronous aware simultaneous orthogonal matching pursuit (AA-SOMP) scheme at central processing unit. This approach is effective to retrieve channel coefficients and active user information of numerous APs, even with restricted pilot length and asynchronous delays. Simulation results show that the proposed approach performs better than the existing algorithms

Mega low Earth orbit (LEO) satellite network has attracted a lot of attention due to its important roles in 6G systems and its many appealing features, including large bandwidth, low latency, extended coverage, and a wide range of applications. However, the mobility of both LEO satellites and terrestrial users necessitates frequent location updates, and a wide satellite coverage adds complexity to its paging signaling. This work proposes a two-layer distributed location management scheme based on dynamic tracking area (TDLMDTA) to reduce the overheads related to paging and location updates. Terrestrial coverage areas are divided into three categories: dynamic tracking areas (DTAs), local registration areas (LRAs), and global registration areas (GRAs). A two-layer DLM architecture is developed by designating the management of GRAs and LRAs, respectively, for global and local location management centers. DTA can adjust its radius dynamically based on the rate of arriving calls and user mobility patterns. The simulation results verify that the proposed TDLM-DTA scheme outperforms traditional location management (LM) schemes in terms of paging delay and LM overhead. This research work focuses on the issues of distributed management and DTA design for advanced LM schemes, laying a foundation for further investigations on dynamic distributed LM in LEO satellite networks.

Deep learning (DL) has been widely utilized in automatic modulation classification (AMC), and its performance depends largely on the presence of high-quality datasets. Motivated by this fact, this work addresses the AMC challenges in data-limited IoT environments, proposing a framework combining few-shot meta-learning and federated learning (FL) for resource-constrained devices, where edge nodes use meta-learning for training with global updates via federated averaging (FedAvg). The system aggregates samples from multiple nodes while still maintaining data security. Simulations involved 11 modulation types with varying SNRs, 100 client nodes, and ten rounds of FL. The iterative process includes loading pretrained parameters, performing local training, averaging local parameters, and updating global parameters. The obtained results show 70% post-training testing accuracy, with a consistently good performance during federated iterations. The results demonstrated the effectiveness of the proposed framework in data-scarce IoT scenarios, offering a robust performance across varying signal qualities while minimizing energy consumption and communication overhead, which is crucial for IoT device longevity and network scalability, highlighting framework's potential for real-world applications in distributed modulation recognition.

With the increasing prevalence of Internet of Things (IoT) devices, security vulnerabilities and malware infections have emerged as significant risks. To address these challenges, advanced vulnerability detection tools are essential for enhancing IoT security assessments. In this study, we analyzed common vulnerabilities and evolving attack methodologies to develop improved detection techniques. Our research focuses on two key areas: 1) comprehensive vulnerability detection and 2) malware infection testing strategies. Through on-site testing and detailed analysis, we identified prevalent security flaws in IoT devices and developed a suite of tools tailored for detecting these vulnerabilities. Additionally, we discovered that some devices exhibit inherent immunity to specific malware strains, emphasizing the need for novel malware infection detection strategies. Real-world evaluations uncovered previously unknown vulnerabilities and weaknesses, revealed widespread susceptibility to DoS attacks, and demonstrated that not all devices are vulnerable to malware infections. These findings confirm the effectiveness of our approach in identifying risks and enhancing IoT security.

Fluid antennas, including those based on liquid, mechanical, and pixel-based technologies, are poised to significantly enhance next-generation wireless systems by adaptively optimizing their radiation characteristics. Many theoretical analyses assumed near-instant reconfiguration, perfect channel knowledge, static or slowly varying propagation environments, and ideal material properties that rarely hold in practice. In this article, we dissect these common assumptions and contrast them with the realities of finite actuation time, limited and imperfect channel state information, rapidly changing fading conditions, electromagnetic coupling, and mechanical constraints. Through illustrative examples and simulations, we demonstrate how ignoring these factors can lead to overestimated gains in capacity, coverage, etc.. We then propose modeling refinements, experimental validation methods, and emerging control algorithms that better account for real-world constraints. Our findings highlight that, while reconfigurable antennas remain highly promising for B5G/6G and Internet of things (IoT) applications, their full potential can only be realized by incorporating practical considerations into system design and performance evaluation.

This letter explores the applications of fluid antenna in wireless semantic communications, aiming to fill up a gap in existing research. In particular, this letter applies Fluid Antenna for Importance-based Ranking (FAIR) systems, which can reconfigure communication encoders and fluid antenna parameters dynamically based on the significance of individual words in a sentence. A reconfiguration and power optimization algorithm is proposed to enhance transmission efficiency. The system's performance is evaluated using Word Error Rate (WER) and ROUGE metrics, demonstrating its potential to improve the performance of semantic communications, particularly in dynamic and resource-constrained environments.

In this paper, we investigate fairness-aware intelligent reflecting surface (IRS)-assisted multiple-user uplink communication systems. Current IRS-assisted communication methods do not consider the fairness issues of communication performance for multiple users. We formulate the IRS phase shift and receive beamforming design as an α-fair utility maximization problem by considering different fairness metrics, including zero fairness and proportional fairness, followed by an iterative algorithm to solve this problem. To overcome the high computational complexity of the iterative algorithm, we introduce a deep learning-based framework, which leverages convolutional neural networks with residual and attention mechanisms for real-time phase shift prediction. Extensive numerical simulations demonstrate that the proposed method achieves performance comparable to the traditional approach while significantly reducing computational overhead, making it a promising solution for future fairness-aware IRS-assisted communication systems.

More effective multiple access techniques are required to fulfill the pressing demand of supporting a large number of connected devices in future mobile communications. This paper proposes an $M$ -ary spread multiple access (MSMA) scheme, designed to provide a high overload ratio for uplink cellular communications. The MSMA scheme distinguishes users by employing codes with superposition constellations. Specifically, it utilizes $M$ -ary code mapping scheme to map each user's data stream onto several subcarriers, thereby achieving a frequency diversity gain. The receiver decodes incoming signals using $M$ -ary code demapping based on generalized sphere decoding (GSD), which effectively balances diversity gain and multi-user interference, leading to enhanced performance. Both analytical and simulation results confirm that the MSMA system yields a lower bit error rate (BER) than power domain non-orthogonal multiple access (PD-NOMA) with the same overload ratio. In addition, MSMA surpasses sparse code multiple access (SCMA) in terms of both BER and overload ratio when the same number of subcarriers are reused. Therefore, with the same number of subcarriers and the same BER requirement, MSMA can support more users than SCMA and PD-NOMA.

Non-orthogonal multiple access (NOMA) is an important multiple access technology for next generation wireless communications. This work focuses on realtime resource allocation in NOMA systems based on reinforcement learning (RL). Q-learning (QL) is an agile RL approach that can adjust its learning strategy to dynamic channel state, making it a perfect machine learning algorithm that can tell agents what to do to maximize its rewards. It does not require a given model of the environment and thus can work adaptively in different scenarios. However, the majority of existing works treated QL as a tool to solve an optimization problem. In this work, QL participates the entire resource allocation process in NOMA systems. As long as user equipment (UE) locations are given, it can optimize resource allocation to achieve a maximum sum rate. In particular, we demonstrate its effectiveness with simulation results in three NOMA schemes, including multi-user superposition transmission (MUST), pattern division multiple access (PDMA), and sparse code multiple access (SCMA) systems. The excellent tracking convergence property of the proposed schemes makes it an ideal choice to perform realtime resource allocation in wireless communications.

Mega low earth orbit (LEO) satellites provide seamless global coverage, which pose a great challenge to location management (LM) due to the high-speed movement of satellites. In this letter, we design a distributed location management (SDLM) scheme with location management centers (LMCs) deployed onboard LEO satellites. We propose an enhanced binary bi-velocity particle swarm optimization (EBBV-PSO) algorithm to solve the LMC deployment problem. Additionally, we introduce a concept of dynamic tracking area (DTA) and derive its optimal radius explicitly. Simulation experiments demonstrate that the proposed SDLM scheme outperforms traditional LM schemes in terms of paging delay and LM overhead.

Cell-free massive multiple-input multiple-output (CFmMIMO) is one of the key enabling technologies for massive machine type communications (mMTC). While CFmMIMO can achieve a macro diversity gain in mMTC, its fronthaul link capacity severely limits system performance. To balance fronthaul overhead and detection/estimation performance, in this paper we investigate joint user activity detection and channel estimation (JADCE) and develop a framework for uplink grant-free CFmMIMO systems based on mixed-quantization fronthaul compression. First, we model the JADCE problem as a mixed-quantization compressing sensing problem, making use of sporadic transmission nature of devices in mMTC. A multiple measurement vector (MMV) generalized approximate message passing (GAMP) algorithm is proposed to improve the performance of activity detection and channel estimation. Unknown priori parameters of the GAMP algorithm can be estimated by an expectation-maximization algorithm. In addition, we design a threshold detection rule at central processing unit (CPU) side and derive a lower bound of active user detection performance in a centralized CFmMIMO system. Numerical results reveal that the proposed mixed-quantization scheme can strike a good balance between fronthaul overhead and detection/estimation performance, verifying that a CFmMIMO with mixed-quantization scheme can perform well in activity detection and channel estimation by selecting an appropriate number of quantization bits.