IEEE Open Journal of the Communications Society最新文献

{"title":"Attention-Enhanced Hybrid Architecture for Efficient Intrusion Detection in Industrial IoT","authors":"Ahmed Elwhishi;Awad A. Younis;Adnan Akhunzada","doi":"10.1109/OJCOMS.2026.3661768","DOIUrl":"https://doi.org/10.1109/OJCOMS.2026.3661768","url":null,"abstract":"The rapid convergence of the Internet of Things (IoT) and the Industrial Internet of Things (IIoT) has increased exposure to advanced and coordinated cyber threats. Existing deep learning-based intrusion detection system (IDS) designs often suffer from high computational cost, limited interpretability, and poor handling of temporal dependencies in IIoT traffic. To address these challenges, we propose CKAN–BiLSTM, a hybrid architecture that integrates 1D convolutional Kolmogorov–Arnold networks (CKAN) for compact and interpretable spatial feature extraction with a bidirectional long short-term memory (BiLSTM) module for temporal modeling and a dot-product attention mechanism for dynamic feature prioritization. The design preserves the parameter efficiency of Kolmogorov–Arnold networks (KANs) while capturing the long-range dependencies essential for IIoT traffic analysis. The proposed framework is evaluated on the publicly available TON-IoT dataset, which includes telemetry and network traffic from diverse IoT and IIoT devices, spanning seven device types and multiple attack families. The model is extensively validated under a 10-fold cross-validation setup using both conventional and advanced performance metrics. Device-wise analysis shows consistently low off-diagonal errors despite class imbalance, and benchmark comparisons situate our results within a broad range of IoT/IIoT IDS studies, indicating competitive performance without relying on parameter-heavy architectures. These results establish CKAN–BiLSTM as an effective and lightweight IDS for real-time IIoT intrusion detection, paving the way for future extensions to streaming adaptation and privacy-preserving federated learning.","PeriodicalId":33803,"journal":{"name":"IEEE Open Journal of the Communications Society","volume":"7 ","pages":"1330-1347"},"PeriodicalIF":6.2,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11373244","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148489716","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An Enhanced Golden Jackal Optimizer for Multi-Criteria DAG Task Scheduling in a Heterogeneous 6G Air–Ground–Space Integrated Network","authors":"Ahmed I. Awad;Marwa M. Khashaba;Mostafa M. Fouda;Zubair Md Fadlullah;Khalid M. Hosny","doi":"10.1109/OJCOMS.2026.3689759","DOIUrl":"https://doi.org/10.1109/OJCOMS.2026.3689759","url":null,"abstract":"Next-generation computing paradigms are redefining the 6G ecosystem by enabling distributed intelligence over an integrated Space–Air–Ground Network (SAGIN) continuum. Despite these advances, scheduling complex applications modeled as Directed Acyclic Graphs (DAGs) within this high-fidelity, three-dimensional infrastructure remains an NP-hard challenge, exacerbated by stochastic mobility, intermittent connectivity, and multi-tenancy resource contention. In this study, we present the Enhanced Golden Jackal Optimizer (EGJO). This training-free metaheuristic jointly optimizes four key objectives: execution time (makespan), local energy consumption, probabilistic task reliability, and operational costs under non-linear congestion-aware pricing. To navigate the non-convex fitness landscape and mitigate premature convergence, a stagnation-aware hybrid strategy is introduced that integrates a Frame Shifting operator for global exploration with a load-balancing intensification heuristic. The performance of EGJO is assessed using high-fidelity simulations that incorporate Gauss-Markov mobility models and LEO orbital constraints. Comprehensive evaluations against six metaheuristics (HHO, WOA, GTO, PSO, GA, and standard GJO) and domain-specific state-of-the-art approaches—including Modified HEFT and Deep Reinforcement Learning (DRL)—demonstrate that EGJO attains a superior Pareto-optimal trade-off and a broader solution spread. Experimental results reveal that EGJO achieves nearly a 33x reduction in makespan compared to GA on large-scale 97-task workloads and reduces handover penalties to 3.0 s, significantly outperforming the 8.5–15.0 s range reported by baseline algorithms. Overall, EGJO achieves up to a 60.5% reduction in the global cost function, with average improvements of 57.67% in makespan and 62.83% in operational costs, ensuring robust and adaptive scheduling in dynamic 6G computing environments.","PeriodicalId":33803,"journal":{"name":"IEEE Open Journal of the Communications Society","volume":"7 ","pages":"6672-6702"},"PeriodicalIF":6.2,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11503280","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148489759","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Detecting Tor Snowflake as Anomalous WebRTC Traffic Under Extreme Class Imbalance","authors":"Ban Alomar;Zouheir Trabelsi","doi":"10.1109/OJCOMS.2026.3705147","DOIUrl":"https://doi.org/10.1109/OJCOMS.2026.3705147","url":null,"abstract":"Tor Snowflake employs the WebRTC protocol stack to circumvent network censorship by simulating conventional peer-to-peer browser communication. Despite modern trained detection algorithms achieving accuracies beyond 99% in laboratory conditions, these evaluations fail to consider the significant class imbalance present in operational contexts where circumvention is infrequent. In these circumstances, the base-rate fallacy induces a significant accuracy decline, resulting in the majority of positive identifications being false positives and making supervised approaches operationally infeasible. This study mitigates these deficiencies by utilizing a 36-dimensional protocol fingerprint vector to identify distinctive patterns in DTLS content types, STUN binding attributes, TLS record distributions, TCP flag configurations, and packet-level entropy. By training exclusively on standard network traffic, the framework recognises Snowflake traffic as structural anomalies without necessitating labeled circumvention samples. A systematic evaluation across realistic operational imbalance ratios indicates that the local outlier factor (LOF) demonstrates strong performance, attaining a 91.6% F1-score (86.0% precision, 98.0% recall) at 1000:1, while supervised classifiers collapse to 17.8–37.6% precision at the same ratio. Deep learning techniques (AE, VAE, GANomaly) attain intermediate performance, with F1-scores ranging from 68.7% to 87.3% at 1000:1.","PeriodicalId":33803,"journal":{"name":"IEEE Open Journal of the Communications Society","volume":"7 ","pages":"6723-6740"},"PeriodicalIF":6.2,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11570227","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148489764","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Temporal Characterization of Fade Events in mmWave Channels","authors":"Claudio Bastos da Silva;Horacio Tertuliano Filho","doi":"10.1109/OJCOMS.2026.3683010","DOIUrl":"https://doi.org/10.1109/OJCOMS.2026.3683010","url":null,"abstract":"Millimeter-wave communication systems operate under propagation conditions markedly different from those observed at lower frequencies, characterized by increased path loss, reduced effective scattering, and a strong dependence on line-of-sight links. Under such conditions, an accurate temporal characterization of fade events becomes essential for assessing link reliability and dynamic performance. This paper investigates the temporal behavior of fading in millimeter-wave channels by analyzing the recurrence and persistence of envelope fluctuations across a range of normalized threshold levels. A modeling approach is adopted in which fast envelope variations are described through a conditional formulation, while slower channel fluctuations are incorporated via a latent-scale mechanism, naturally accommodating uncertainty through a Bayesian modeling perspective. The resulting envelope representation enables a consistent evaluation of temporal fading metrics without relying on derivative-based assumptions. Analytical and semi-analytical formulations are obtained for key quantities associated with fade events, including the cumulative behavior of the envelope, the frequency of level crossings, and the average duration of fades. Numerical results based on Monte Carlo simulations are presented and compared with commonly adopted fading models, allowing a direct assessment of the predicted temporal dynamics under millimeter-wave propagation conditions. The results demonstrate that the proposed formulation captures the combined effects of rapid signal fluctuations and slower channel variability, providing a physically consistent basis for the temporal characterization of fade events in high-frequency wireless channels.","PeriodicalId":33803,"journal":{"name":"IEEE Open Journal of the Communications Society","volume":"7 ","pages":"3813-3831"},"PeriodicalIF":6.2,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11479184","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148489765","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"DyQRA: A Mathematical Framework for Enhanced Dynamic QoS Provisioning in 5G/6G Networks","authors":"Mostafa Panahandeh;Vahid Shah-Mansouri;Mohammad Reza Pakravan;Seyed Hamed Rastegar","doi":"10.1109/OJCOMS.2026.3652589","DOIUrl":"https://doi.org/10.1109/OJCOMS.2026.3652589","url":null,"abstract":"The increasing complexity of 5G and emerging 6G networks necessitates advanced resource allocation and adaptive scheduling mechanisms to ensure stringent Quality of Service (QoS) requirements. Existing heuristic solutions struggle to scale with the high dimensionality of next-generation networks, leading to sub-optimal packet scheduling and queue management. This paper proposes a mathematical optimization framework, termed DyQRA (<underline>Dy</u>namic <underline>Q</u>oS-Aware <underline>R</u>esource <underline>A</u>llocation), that integrates active queue management with dynamic scheduling, ensuring queue stability, minimal delay, and compliance with Packet Delay Budget (PDB) and Packet Error Rate (PER) constraints. Our model leverages Lyapunov optimization, specifically the drift-plus-penalty technique. Through frame-based modeling, our method effectively captures time-varying traffic dynamics and service-rate fluctuations, ensuring robust QoS adherence. Additionally, the model accounts for effective arrival rates, considering packet drops caused by delay violations. Simulation results demonstrate that the proposed method outperforms existing schemes, making it a strong candidate for implementation in both 5G and 6G networks.","PeriodicalId":33803,"journal":{"name":"IEEE Open Journal of the Communications Society","volume":"7 ","pages":"701-717"},"PeriodicalIF":6.2,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11345477","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148491133","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Iterative Joint Target and Covert Data Detection Algorithm for FD-ISAC","authors":"Cuong Huy Huynh;Nam Quoc Le;Hung Nguyen Manh;Lap Luat Nguyen;Le-Nam Tran;Hien Quang Ta","doi":"10.1109/OJCOMS.2026.3693923","DOIUrl":"https://doi.org/10.1109/OJCOMS.2026.3693923","url":null,"abstract":"We investigate a full-duplex integrated sensing and communication system (FD-ISAC) where a base station (BS) simultaneously detects moving targets and receives covert uplink data. In such a system, self-interference (SI) from co-located antennas and reflections from static and slow-moving objects can severely degrade both sensing and communication performance. To address this critical issue, we propose an iterative frequency-domain algorithm for joint target and data detection. In each iteration, the estimated target echoes and covert communication signals are removed to obtain an improved estimate of the Doppler deviation, which is used to suppress SI via a smoothing technique. With SI significantly reduced, a simple radar processing method, such as the two-dimensional (2D) periodogram, can be effectively applied to the residual signal for target detection. In the final step, data detection is performed after the effects of SI and target echoes have been largely mitigated. Numerical results show that the proposed approach outperforms the best-known solution, particularly in the presence of slow-moving interference. Notably, it achieves near-optimal sensing and communication performance with only two iterations and a small number of pilot symbols, requiring only a 0.5 dB margin, compared to the ideal communication-only case. Under the low-power regime of covert transmissions, a single pilot symbol is sufficient to achieve near-optimal sensing and communication performance while maintaining the best covertness performance. These results highlight the practical potential of the proposed algorithm for FD-ISAC implementations with and without covertness considerations.","PeriodicalId":33803,"journal":{"name":"IEEE Open Journal of the Communications Society","volume":"7 ","pages":"6012-6023"},"PeriodicalIF":6.2,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11520822","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148491555","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Ultra-Efficient and Scalable QoS Prediction for QKD in 7G Networks Using Fuzzy-Augmented Neural Networks","authors":"Francesco Fiorini;Cristian Bua;Rosario G. Garroppo;Michele Pagano;Davide Adami;Stefano Giordano","doi":"10.1109/OJCOMS.2026.3715201","DOIUrl":"https://doi.org/10.1109/OJCOMS.2026.3715201","url":null,"abstract":"The integration of Quantum Key Distribution (QKD) into prospective 7G communication systems represents a promising approach to provide information-theoretic security for future networked applications. In such environments, QKD infrastructures are expected to support heterogeneous services with diverse security and performance requirements under strict resource constraints. However, QKD performance is inherently sensitive to dynamic channel conditions, which directly affect key metrics such as the Quantum Bit Error Rate (QBER) and the Secret Key Rate (SKR). Since these parameters determine the achievable Quality of Service (QoS), their accurate and efficient prediction is essential for effective system optimization. In this work, we propose an ultra-efficient QoS prediction framework for QKD systems based on a Fuzzy-Augmented Neural Network (FANN) architecture. The proposed approach decomposes the prediction task into a regression stage, followed by an application-specific fuzzy inference layer that maps predictions into QoS classes. This decoupled design enables a lightweight, modular, and extensible solution suitable for resource-constrained QKD nodes. The proposed framework supports dynamic quantum channel scheduling, allowing applications to access key generation resources according to their security and throughput requirements, with direct benefits in terms of energy efficiency at both control and physical layers. Experimental results on a publicly available QKD dataset demonstrate that the proposed approach achieves comparable or improved <inline-formula> <tex-math>$F_{1}$ </tex-math></inline-formula>-score performance with respect to a conventional Feed-Forward Neural Network (FFNN) classifier, while reducing inference time by over 98% and memory usage by more than 71%, even compared to FFNN models that leverage classical compression techniques, i.e., pruning and quantization.","PeriodicalId":33803,"journal":{"name":"IEEE Open Journal of the Communications Society","volume":"7 ","pages":"8620-8632"},"PeriodicalIF":6.2,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11614818","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148628178","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Adaptive Weight Gating: A Performance Enhancement Method for Analog-Digital Cascaded Self-Interference Cancellation","authors":"Jiaqi Liang;Yaxing Li;Jinling Xing;Yufeng Ma;Fangmin He;Jin Meng","doi":"10.1109/OJCOMS.2026.3715198","DOIUrl":"https://doi.org/10.1109/OJCOMS.2026.3715198","url":null,"abstract":"With the high-density integration of electronic systems on modern integrated platforms, co-site high-power electromagnetic interference has become a critical bottleneck degrading electromagnetic compatibility of such platforms. Analog-digital cascaded self-interference cancellation (SIC) is widely recognized as a promising solution, but its performance is limited by the dynamic coupling between the analog and digital stages. In this paper, a unified time-domain model is constructed for the analog-digital cascaded SIC system consisting of the analog least mean square (ALMS) loop and digital multi-tap cancellation. Rigorous theoretical derivation demonstrates that the additional alternating current (AC) components induced by adaptive analog weight variation lead to irreducible residual interference and signal coupling, which fundamentally limit the overall performance of the cascaded system. To address this issue, an adaptive weight gating method is proposed, which accurately determines the enable timing of the digital stage, and adapts to time-varying multipath scenarios. By judging the power intensity of the input signal in the digital SIC circuit and the variation rate of analog weight, this method controls whether to enter the weight adaptation phase or the weight fixation phase, thereby eliminating extra AC components. Both simulation and experimental results demonstrate that the proposed method improves the overall interference cancellation ratio (ICR) by more than 30 dB in the fixation phase compared with the adaptation phase, under a typical co-site scenario with 20 MHz bandwidth, 20 dBm transmit power, and a multipath channel with four taps. This study provides a practical and high-performance solution for the deployment of cascaded SIC technology.","PeriodicalId":33803,"journal":{"name":"IEEE Open Journal of the Communications Society","volume":"7 ","pages":"9104-9113"},"PeriodicalIF":6.2,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11614899","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148748041","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Construction of Hirota Bilinear Neural Network Layer for Deep Learning-Based Classification of Wireless Communication Signals","authors":"Nguyen Minh Tuan;Quang-Sang Nguyen;Nguyen van Mui","doi":"10.1109/OJCOMS.2026.3718378","DOIUrl":"https://doi.org/10.1109/OJCOMS.2026.3718378","url":null,"abstract":"Deep learning has become an essential tool for analyzing complex patterns in wireless communication signals; however, conventional architectures such as SimpleRNN, LSTM, and Transformer-based networks often show limited capability in capturing highly nonlinear feature interactions and global dependencies simultaneously. To address this limitation, this paper proposes a novel Hirota Bilinear Neural Network (Hirota-BiNN) framework inspired by Hirota’s bilinear formalism from nonlinear mathematical physics. The proposed model introduces a Hirota bilinear layer that explicitly models multiplicative interactions between two parallel feature branches through full bilinear mapping and low-rank bilinear refinement, thereby improving representation capacity while maintaining computational efficiency. The new layer is constructed for deep learning and published at <uri>https://pypi.org/project/blinear/</uri> with version 1.0.0. Real-world wireless communication signals from Wi-Fi, LTE, and 5G-NR systems are employed for binary classification tasks. Raw in-phase and quadrature (IQ) signals are transformed into magnitude sequences, segmented into fixed-length samples, and labeled using an energy-based thresholding strategy. For fair evaluation, SimpleRNN, LSTM, Transformer, and the proposed Hirota-BiNN are trained under identical experimental settings. Experimental results demonstrate that conventional recurrent and attention-based models achieve only moderate performance, with testing accuracy around 69.9% and limited precision, recall, and F1-score due to insufficient nonlinear feature modeling. In contrast, the proposed Hirota-BiNN achieves significantly faster convergence, stronger classification stability, and superior predictive performance by explicitly learning nonlinear multiplicative dependencies between signal features. The corrected and reproducible implementation is further released as the publicly available Python package <monospace>blinear</monospace> to ensure experimental transparency and practical deployment. These findings confirm that mathematically inspired bilinear neural structures provide a promising direction for intelligent wireless signal analysis and future physics-informed deep learning systems.","PeriodicalId":33803,"journal":{"name":"IEEE Open Journal of the Communications Society","volume":"7 ","pages":"9034-9050"},"PeriodicalIF":6.2,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11630612","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148748444","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-Agent DRL for Cooperative Resource Allocation in C-NOMA-Enabled Multi-UAV Networks With 2-D Hybrid Beamforming","authors":"Muhammet Hevesli;Abegaz Mohammed Seid;Mohamed Abdallah;Aiman Erbad","doi":"10.1109/OJCOMS.2026.3717101","DOIUrl":"https://doi.org/10.1109/OJCOMS.2026.3717101","url":null,"abstract":"Sixth-generation (6G) wireless networks require massive connectivity, high spectral efficiency (SE), and energy efficiency (EE). Although conventional non-orthogonal multiple access (NOMA) improves spectrum utilization by allowing multiple users to share the same subchannel through power-domain multiplexing, applying NOMA to large user groups significantly increases successive interference cancellation (SIC) complexity and intra-group interference. To address this limitation, clustered NOMA (C-NOMA) groups users into small subclusters where SIC is performed over fewer users, thereby improving scalability while reducing decoding complexity. Combining C-NOMA with hybrid beamforming (HB) further enhances SE and lowers power consumption by serving each subcluster through a dedicated analog beam with fewer radio-frequency chains. In this paper, we develop a cooperative resource-allocation framework for C-NOMA-enabled multi-uncrewed aerial vehicle (UAV) millimeter-wave (mmWave) networks with two-dimensional (2D) HB. UAVs equipped with 2D uniform planar array antennas serve as aerial base stations to enhance line-of-sight (LoS) connectivity. We formulate an EE-maximization problem that jointly considers user subclustering, power allocation (PA), and subchannel assignment (SA) under transmit-power and SIC constraints. User subclustering is first performed through head selection and channel-disparity-based pairing to enable efficient C-NOMA transmission. Given the resulting structure, the 2D HB and joint PA/SA optimization problem is solved via a two-stage approach: a heuristic 2D HB construction followed by a multi-agent deep deterministic policy gradient (MADDPG) algorithm for cooperative PA and SA optimization under centralized training and decentralized execution (CTDE). Simulation results demonstrate that the proposed framework consistently improves EE over representative benchmarks across different network configurations.","PeriodicalId":33803,"journal":{"name":"IEEE Open Journal of the Communications Society","volume":"7 ","pages":"8998-9016"},"PeriodicalIF":6.2,"publicationDate":"2026-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11625804","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148747176","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}