The Internet of Underwater Things (IoUT) supports marine sensing, environmental monitoring, subsea inspection, and autonomous underwater operations. However, IoUT communication is constrained by limited bandwidth, long propagation delay, time-varying underwater channels, intermittent connectivity, and strict energy budgets. Semantic Communication (SC) offers a promising alternative by transmitting task-relevant meaning rather than raw data, thereby improving communication efficiency in resource-constrained underwater networks. This paper presents a critical and feasibility-aware survey of SC for IoUT, focusing on opportunities, challenges, limitations, and future research directions. We first review the fundamentals of SC-enabled IoUT systems, including semantic representations, layered architectures, semantic channel modeling, and task-oriented evaluation metrics. We then examine learning-driven approaches based on machine learning (ML), knowledge graphs (KGs), vision-language models (VLMs), generative models, and federated learning (FL), with emphasis on their feasibility under underwater edge constraints. Representative applications, including environmental monitoring, marine ecology, subsea infrastructure inspection, disaster response, and autonomous underwater vehicle (AUV) coordination, are analyzed from an SC perspective. Finally, we identify key research directions involving standardized semantic models, reproducible testbeds, compute–communication trade-offs, trustworthy reconstruction, hybrid underwater links, energy-aware edge intelligence, semantic security, digital twins (DTs), and cross-domain interoperability. This survey provides a structured foundation for developing reliable, efficient, and meaning-driven IoUT communication systems.

Beyond-line-of-sight (BLOS) communication is a fundamental component of modern defense strategies, enabling secure and reliable information exchange in environments where traditional line-of-sight (LOS) methodologies are ineffective due to obstructions or operational constraints. This article explores the key technologies advancing BLOS communication, with a focus on nanonetworks, aerial relays, and satellite-based defense communication. To underscore the real-world applications and challenges of BLOS systems, we present two practical case studies, one examining tropospheric ducting for BLOS maritime communication, highlighting its advantages and implementation challenges, and another on UAV path planning in radar-threat war zones, demonstrating the role of optimization techniques in enhancing operational efficiency. Furthermore, this study identifies several critical future research directions in BLOS defense communication, including resilience enhancement, heterogeneous network integration, contested spectrum management, advancements in multimedia communication, adaptive methodologies, and the expanding domain of the Internet of Military Things (IoMT). Addressing both technological advancements and real-world constraints, this research lays a foundation for strengthening BLOS communication systems, fostering cross-sector collaboration, and driving innovation at the intersection of defense, academia, and industry.

Terahertz (THz) beamspace massive multiple-input multiple-output (MIMO) systems exploit lens-based beamforming to reduce radio-frequency (RF)-chain complexity and capitalize on angular sparsity. Under mobility, however, beam directions drift, making exhaustive beam training and per-slot sparse recovery both training intensive. Prior-aided (PA) beam tracking reduces this burden by probing a local window around a predicted direction, but a fixed window is brittle because it ignores time-varying prediction uncertainty. To address this limitation, we propose an uncertainty-aware tracking framework comprising KPA-BT and its adaptive-window extension KPA-BT-AV. The framework couples Kalmanbased state prediction and uncertainty quantification with circular local beam probing, hybrid peak/centroid measurement extraction, and covariance-driven window adaptation. A coverage-constrained analysis further links the predicted covariance to the minimum nominal window size under a Gaussian beam-index error model, providing analytical guidance for the practical clipped controller used in simulation. Using an off-grid THz channel model, simulations benchmarked against OMP, CoSaMP, fixed-window PA, and a lightweight learning-enhanced baseline show that KPA-BT improves channel-estimation NMSE, while KPA-BT-AV increases net sum-rate by reducing training overhead without sacrificing near-ceiling throughput in training-limited settings. Additional evaluations under multipath blockage, abrupt maneuvers, wideband beam squint, and joint (N,K) scaling further characterize robustness, practical limits, and the conditions under which richer multi-cluster or wideband models become necessary.

The rapid evolution from 5G toward 6G has intensified the demand for energy-efficient communication systems capable of supporting massive connectivity, ultra-low latency, and ubiquitous coverage. Conventional energy efficiency (EE) metrics, such as bits-per-Joule, are insufficient for capturing the unique characteristics of passive and semi-passive paradigms, including backscatter communications (BackCom), where energy harvesting, sporadic transmissions, and circuit-level inefficiencies dominate. This paper introduces and examines the waste factor (W) as a unified metric for evaluating energy inefficiency in both active and passive wireless systems. W quantifies the ratio of total consumed power to useful output power, enabling deeper insight into wasted energy across devices, circuits, and cascaded networks. The review outlines theoretical foundations, demonstrates W’s applicability to monostatic, bistatic, and ambient backscatter architectures. Comparative analysis shows that conventional EE metrics may yield misleading results, while W provides a more realistic evaluation of energy usage, revealing trade-offs in throughput, circuit losses, and system scalability. Furthermore, the paper highlights how the W can guide energy-aware design in hybrid architectures such as Reconfigurable Intelligent Surface (RIS)-assisted backscatter systems, Multiple-Input Multiple-Output (MIMO) deployments, and Artificial Intelligence (AI)/Large Language Model (LLM)-driven optimization frameworks. The W thus establishes itself as a holistic tool for benchmarking green wireless technologies, with significant implications for Internet of Things (IoT) applications, sustainable sixth-generation (6G) networks, and future standardization efforts.

Reconfigurable intelligent surfaces (RISs) can improve coverage and energy efficiency, but their practical value depends on reliable initial access before alignment is established. In that phase, the training signal-to-noise ratio (SNR) at the user can be low, making hierarchical and hard-decision coded beam training vulnerable to error propagation. This paper develops confidence-aware layered matching (CALM), a receiver-centric beam training framework that extracts soft reliability information from the four beam-tuple measurements collected at each training layer. CALM-JD (joint decoding) performs reliability-weighted decoding of the BS (base station) and RIS hypotheses from the shared layer measurements instead of committing to brittle layer-wise hard decisions. CALM-APA (adaptive pilot allocation) further reallocates extra pilots only to the least reliable layers, while CALM-APA+KCONF (top-K confirmation) adds a bounded top-K tuple-confirmation stage to improve robustness without losing runtime predictability. The resulting methods preserve the low-overhead layered structure of coded beam training while addressing low-SNR reliability and bounded online latency. Complexity analysis and relative software runtime profiling are reported alongside achievable-rate and misalignment results. For a 28-GHz BS–RIS–user link with a 64-antenna BS and a 16×16 RIS, the dominant-path baseline shows that the proposed CALM variants significantly reduce misalignment probability in the low-SNR regime and, in the simplified narrowband dominant-path setting studied here, approach the exhaustive-search rate ceiling with roughly two orders-of-magnitude fewer pilots than exhaustive search. Additional narrowband robustness tests with richer sparse multipath (up to eight physical paths per hop), RIS phase quantization, and angular off-grid mismatch preserve the same qualitative ordering, with CALM-APA+KCONF retaining the strongest rate under the tested non-idealities. These results support confidence-aware coded beam training as a physical-layer access mechanism whose pilot savings can be mapped to concrete IoT/CPS (cyber-physical systems) benefits—including reduced per-device access latency and increased device density under fixed frame budgets—while broader wideband, hardware, and system-level validation remains future work.

This work introduces a comprehensive control technique that tackles the significant issue of attitude control in underactuated quadrotor UAVs, particularly accounting for input latency, an often neglected yet vital component in practical applications. A barrier Lyapunov function (BLF) is incorporated to guarantee accurate trajectory tracking and uphold error limits, hence enhancing system stability by restricting the tracking error within a specified range. Furthermore, to alleviate the detrimental impact of input latency on system performance, the suggested framework employs an intermediate variable strategy in conjunction with a Fuzzy Padé approximation technique. This control method markedly improves trajectory precision and system resilience, rendering it ideal for sustainable and mission-critical UAV missions. The efficacy of the method is substantiated by simulation outcomes and further corroborated by hardware implementation on a 3-degree-of-freedom (DOF) hover system by Quanser, ensuring congruence between software and actual performance.

The integration of Artificial Intelligence (AI) and emerging 6G networks introduces new opportunities for scalable coordination in tactical autonomous vehicle systems. This paper proposes a communication-centric hierarchical architecture for Tactical Autonomous Defense Vehicle Networks (TADVNs) that models the integration of edge-assisted Large Language Model (LLM) reasoning with 6G-enabled connectivity and semantic communication. The framework is designed to improve coordination efficiency, reduce communication overhead, and enhance latency resilience under increasing fleet-scale operation. Unlike conventional task-specific AI pipelines that rely on structured feature processing and rule-based coordination, the proposed approach incorporates semantic abstraction and context-aware decision support within a layered edge-cloud communication architecture. We evaluate communication and coordination performance via Monte Carlo simulations across fleet sizes of 5-30 vehicles under contested network conditions. Results indicate that at 30-vehicle scale, the 6G-LLM configuration achieves 75.2% latency reduction (29.1 ms versus 117.5 ms), a 68.7 percentage point increase in mission success rate (82.9% versus 14.2%), and an 88.6% reduction in communication overhead compared to a 5G-based conventional AI baseline. These findings demonstrate measurable benefits in coordination and communication when semantic reasoning is combined with low-latency 6G connectivity.

The integration of UAVs and Simultaneous Transmitting and Reflecting Reconfigurable Intelligent Surfaces (STAR-RIS) has emerged as a promising solution for next-generation wireless networks, particularly in disaster and infrastructure-limited scenarios. UAVs provide flexible deployment, while STAR-RIS improves coverage and fairness by enabling concurrent signal reflection and transmission. This paper proposes a digital twin-enabled Multi-Relay UAV-assisted STAR-RIS (MR-UAV-STAR-RIS) framework for intelligent rate management, user fairness, and load balancing in Small-Cell (SC) cellular networks. Dual-connected user equipment simultaneously accesses the Macro-Cell (MC) via MR-UAV-STAR-RIS and SC base stations over mmWave links. A joint optimization problem is formulated involving active beamforming at the MC and SC-BSs, passive STAR-RIS/RIS beamforming, and 3D UAV positioning. To address the high-dimensional complexity, a digital twin-enabled multi-task deep reinforcement learning model based on DDPG is developed. Simulation results confirm that the proposed approach significantly improves fairness, load balancing, and overall network performance compared to benchmark schemes under dynamic conditions.

The rapid expansion of satellite mega-constellations and demand for low-latency connectivity pose significant operational challenges that traditional, static approaches cannot address. This paper asserts that the Artificial Intelligence of Things (AIoT) paradigm is the essential, transformative framework for future space communications, fundamentally integrating intelligent processing across the entire physical space and ground segment infrastructure. Moving beyond standard synthesis, this work provides a leading-edge unified AIoT taxonomy for satellite systems derived from a systematic examination of 100 recent publications. This work offers critical insights into the practical implementation and synergistic effects of AIoT across key applications in modern satellite networks, including ground station scheduling, dynamic network optimization, predictive maintenance, and physical security. The insights derived from this work demonstrate how the convergence of distributed sensing, intelligent analytics, and autonomous actuation transforms operations across space, ground, and link segments, a perspective often fragmented in the existing literature. This work highlights the unique utility of AIoT in enabling real-time detection of orbital debris and system interruptions. Furthermore, this work provides critical research frontiers that must be prioritized, addressing the multi-level optimization problem for extreme conditions, the lack of representative training datasets, and the engineering of robust, scalable security protocols against an expanding attack surface. By consolidating these applications and focusing on actionable future development paths, this paper serves as an essential strategic reference for researchers and professionals developing autonomous, resilient, and highly efficient space infrastructure.

Deep Space Communication (DSC) is a critical enabler for reliable data exchange between Earth-based infrastructure and spacecraft operating beyond lunar orbit. This paper presents a comprehensive and up-to-date survey of DSC systems, encompassing architectural foundations, enabling technologies, and emerging research challenges. In particular, the structure and operation of Deep Space Communication Networks (DSCNs) are examined, highlighting the functional interactions among deep space stations, communication complexes, signal processing centers, and mission control centers under severe propagation delays and intermittent connectivity. The paper provides a systematic review of traditional Radio Frequency (RF) communication and emerging Free-Space Optical (FSO) technologies, including hybrid RF/FSO architectures, and analyzes their trade-offs in terms of robustness, bandwidth efficiency, power consumption, and operational complexity. Recent mission demonstrations and international technology roadmaps are discussed to illustrate the ongoing transition toward high-capacity optical links for future lunar, Martian, and deep-space missions. Furthermore, advances at the physical and link layers are surveyed, covering modulation techniques, forward error correction schemes, and adaptive link optimization, with particular emphasis on Low-Density Parity-Check (LDPC) codes, Polar codes, hybrid forward error correction (FEC) schemes, and Adaptive Coding and Modulation (ACM) for operation under low signal-to-noise ratios and time-varying channels. At the networking layer, the paper reviews Consultative Committee for Space Data Systems (CCSDS) standards and Delay/Disruption-Tolerant Networking (DTN) protocols, identifying key open research challenges related to scalability, routing, buffering, quality-of-service support, and autonomous operation. By integrating physical-layer technologies, networking protocols, and system-level considerations, this work outlines emerging trends, including AI-native communication architectures, that are expected to shape the design of scalable, autonomous, and interoperable interplanetary communication networks.

Visible light communication (VLC) channel impairments introduced considerable limitations on designing high-speed data rate communication systems. Thus, many traditional and machine learning (ML)-based equalization techniques have been considered to mitigate linear, nonlinear, constant, and varying distortions. Although ML-based equalizers offer powerful potential to equalize light-emitting diode (LED) nonlinearities, inter-symbol interference (ISI), multi-path fading, and other channel impairments, and predict their complex behavior compared to traditional digital signal processing (DSP)-based equalizers, they require huge computational resources in both their training and model deployments. Moreover, these models typically are trained on and utilized for a single data rate, making them inefficient in real-life situations in which the data rate varies. We developed and experimentally validated post-equalization approaches: artificial neural network (ANN) with 3 hidden layers and a novel binary neural tree (BNT) architecture of depth 4, enabling reliable transmission up to 1Gbps using a single red commercial LED of raw bandwidth of 8.3MHz supporting multi-data rates from 50 to 500 Mbaud. The proposed ML-based post-equalization models are trained on 4-level pulse amplitude modulation (4-PAM) pre-equalized multi-rate signals ranging from 50to 500Mbaud, and evaluated using regression metrics [mean squared error (MSE) and mean absolute error (MAE)] for training and digital communication performance such as bit error rate (BER), eye diagram, and confusion matrix. We showed that the BNT model achieves a 54.23% and 54.37% reduction in resource usage compared to ANN for training and deployment, respectively. To the best of our knowledge, this is the first study to apply BNT equalization in VLC, paving the way for more computational-aware commercial VLC systems.

Foundation models are emerging as a transformative paradigm for AI-native sixth-generation (6G) wireless networks by enabling scalable, transferable, and data-efficient intelligence across diverse communication tasks. Unlike conventional deep learning models that are trained for individual applications, wireless foundation models (WFMs) learn generalized representations from large-scale heterogeneous wireless data and can be efficiently adapted to communication, sensing, localization, and network optimization tasks with minimal task-specific supervision. Despite rapid progress, current research remains fragmented across architectures, training paradigms, and application domains, with no unified survey dedicated to the design, learning, and deployment of WFMs. This survey presents a comprehensive and unified review of wireless foundation models. We first establish the fundamental concepts of WFMs and introduce a taxonomy that organizes the field according to model architectures, pre-training paradigms, and applications. We then review representative architectures, self-supervised pre-training strategies, parameter-efficient adaptation methods, datasets, benchmarks, and evaluation methodologies, highlighting their roles in enabling transferable wireless intelligence. Furthermore, we examine emerging applications spanning physical-layer signal processing, network intelligence, and cross-layer optimization, and discuss the key challenges of data availability, generalization, interpretability, efficient edge deployment, and standardization. Finally, we outline future research directions toward scalable, trustworthy, and general-purpose wireless intelligence for AI-native 6G networks. This survey provides a comprehensive reference for researchers and practitioners developing next-generation intelligent wireless systems.

Electromagnetic (EM) communication is approaching fundamental physical and thermodynamic limits, where further performance gains through spectrum expansion and waveform optimization alone are increasingly unsustainable. The purpose of this paper is to explore how wireless communication may evolve beyond the EM paradigm by reframing information transfer as controlled manipulation of physical, biological, and cognitive states rather than radiative signal propagation. The main contribution of this work is a state-centric conceptual framework for post-6G communication. The paper identifies and categorizes ten foundational paradigms, including quantum-state transfer, atomic and lattice-level signaling, biological communication, cognitive telepresence, and spacetime-based coordination, defining potential non-EM and hybrid communication mechanisms. In addition, a research roadmap is outlined to place these paradigms within plausible future network generations beyond 6G. The key findings of this study are conceptual. The analysis shows that diverse communication mechanisms across physical, biological, and cognitive domains can be unified using common principles such as state transduction, coherence preservation, entropy management, and energy-aware conversion. These findings indicate that future communication systems may evolve from spectrum-bound infrastructures into adaptive and self-organizing networks that integrate information transfer with sensing, computation, and actuation. This work establishes a conceptual reference framework for future theoretical and interdisciplinary research on communication beyond conventional EM-based systems.

Future wireless networks (WNs) must address unprecedented challenges in resource allocation (RA) driven by dynamic environments, diverse user demands, and heterogeneous service requirements. Emerging services such as enhanced mobile broadband (eMBB), ultra-reliable low-latency communications (URLLC), and massive machine-type communications (mMTC) demand intelligent, context-aware resource management strategies beyond traditional methods. Semantic communication (SemCom), which prioritizes conveying intended meaning rather than raw data, offers a promising paradigm to enhance spectral efficiency, reduce communication overhead, and improve user satisfaction. In parallel, advancements in artificial general intelligence (AGI) and large language models (LLMs) introduce new capabilities in reasoning, semantic inference, and adaptive decision-making. This paper presents a unified conceptual and architectural framework that integrates SemCom, LLMs, and AGI for intelligent RA in future WNs. We first examine foundational concepts, then classify and compare methodologies across key performance metrics, and finally explore synergistic architectures that combine these technologies. We highlight open challenges, including semantic metric design, real-time AGI adaptation, scalable LLM deployment, and privacy-preserving semantic reasoning. Unlike prior works, this survey uniquely bridges the semantic and cognitive dimensions of RA, providing a comprehensive roadmap for building fully autonomous, semantic-aware, and resource-efficient wireless communication systems.

UAVs are poised to play a pivotal role in next-generation communication networks, supporting applications from emergency coverage to large-scale IoT connectivity. However, enabling intelligent and autonomous UAV networking remains challenging because conventional Artificial Intelligence (AI) approaches have inherent limitations. Neural models provide adaptability but lack interpretability, while symbolic methods ensure explainability yet struggle with scalability in dynamic environments. This article examines the potential of Neuro-Symbolic Artificial Intelligence (NSAI) to bridge this gap by combining data-driven learning with logic-based reasoning. We outline the fundamentals of NSAI, explain its architectural integration into UAV communication systems across edge and cloud layers, and discuss its capacity to support interpretable, constraint-compliant decision-making in safety-critical missions. Representative use cases illustrate how NSAI can strengthen reliability, adaptability, and transparency in UAV operations. Finally, key challenges and open research directions, such as scalability, adversarial robustness, and alignment with Sixth-Generation (6G) and Non-Terrestrial Networks (NTN), are highlighted, offering a guideline for advancing NSAI as a pathway toward intelligent, resilient, and explainable aerial networks.

This article presents a comprehensive performance evaluation of signal detection algorithms for reconfigurable intelligent surface (RIS)-assisted cell-free massive MIMO systems (CF-mMIMO) in 5G/6G networks. We investigate six detection schemes: zero-forcing (ZF), minimum mean square error (MMSE), conjugate gradient (CG), Neumann-series MMSE, orthogonal approximate message passing (OAMP), and a novel hybrid successive over-relaxation MMSE (SOR-MMSE) detector. The proposed algorithm is specifically designed for diverse system loading conditions, employing a hybrid approach that adaptively switches between SOR iterations for overdetermined systems and residual-based updates for underdetermined systems. It further incorporates adaptive constellation projection and reliability-based soft decisions, optimized for RIS-enhanced environments. We provide mathematical derivations, convergence analysis with eigenvalue-based bounds, and relaxation parameter optimization. Through extensive Monte Carlo simulations encompassing balanced, overloaded, and underloaded antenna-to-user configurations with a 64-element RIS, we demonstrate significant performance improvements. The hybrid SOR-MMSE detector achieves remarkable gains, particularly in underloaded RIS-assisted systems where it provides approximately 2-3 dB signal-to-noise ratio (SNR) improvement compared to ZF detection for an equivalent bit error rate (BER). For balanced configurations, it maintains a 1-2 dB advantage in high-SNR regimes compared to the standard MMSE detector with RIS assistance. The benefits of RIS assistance are evident across all scenarios. For example, an underloaded system reaches a capacity of nearly 19 bps/Hz at 20 dB SNR, which is a significant increase from the 15 bps/Hz achieved without RIS.