• 学术搜索
  • 科研智能体
    • Research Labs
    • AI 阅读
    • AI 文库
    • 深度研究
    • 学者亮点
  • 学术资源
    • AI2000
    • 期刊/会议
    • 学者库
    • 学术API
      MCP&Skill
    • 溯源树
    • 数据集
  • 知识沉淀
    • 学术空间
订阅小程序
旧版功能
aminer vip
开通会员低至0.73元/天
一次搞定AI科研
立即登录
  • English
  • 联系方式
    Annals of Telecommunications

    Annals of Telecommunications

    JournalISSN 0003-4347eISSN 1958-9395中科院 计算机科学 4区

    年发文量

    研究主题

    论文(5946)

    排序
    1Resource-efficient PDCCH Scheduling for Massive IoT in Next-Generation Cellular Networks
    Tiago P. C. de Andrade,Matheus J. Ota,Nelson L. S. da Fonseca,Flávio K. Miyazawa

    The exponential growth of the Internet of Things (IoT) and massive Machine-type Communications (mMTC) applications has introduced unprecedented traffic demands in next-generation (NG) cellular networks. Crucially, while a user-plane scheduler may successfully allocate Physical Resource Blocks (PRBs) on data channels (Packet Uplink Shared Channel (PUSCH)/Physical Downlink Shared Channel (PDSCH)), transmission cannot occur without successful delivery of Downlink Control Information (DCI) over the Physical Downlink Control Channel (PDCCH). Consequently, unallocated control requests render user-plane decisions void, positioning the PDCCH as a primary system bottleneck. Finite control resources and rigid mapping constraints frequently cause message blocking and resource stranding, even when data channel capacity is available. To resolve this NP-hard candidate selection problem, this paper formally defines the PDCCH Allocation Problem PDCCH-AP and introduces two Integer Linear Programming (ILP) models, Weighted Control Request Packing (W-CoRP) and Minimum Blocking Control Request Packing (MB-CoRP), to establish theoretical optimality benchmarks. We also propose Global-Aware Probabilistic Packing (eGPP), a novel heuristic that leverages probabilistic methods and multi-core processing to navigate the vast solution space and escape local optima. Extensive numerical simulations demonstrate that eGPP significantly outperforms state-of-the-art baselines by drastically reducing control channel blocking, maximizing throughput, and guaranteeing strict Quality of Service (QoS) and proportional fairness in dense NG deployments. To perform the simulations, we developed a 3rd Generation Partnership Project (3GPP) cellular network-based PDCCH simulator centered on the Medium Access Control (MAC) layer logic and scheduling performance. The simulator faithfully reproduces 3GPP-compliant control channel structures — including control resource allocation, aggregation levels, and Search Space constraints — while employing link-level physical PHY abstraction to evaluate scheduling decisions with maximum protocol fidelity and high computational efficiency.

    2026
    引用
    AI阅读
    加入学术空间
    2Subspace-based Approach for Multiuser Uplink Channel Estimation and Localization Using Extremely Large Antenna Arrays
    Rafik Guellil,Karim Abed-Meraim,Adel Belouchrani

    This paper investigates multiuser uplink channel estimation and localization for Extremely Large Antenna Arrays (ELAAs), explicitly accounting for multipath propagation and intersymbol interference (ISI). Multiple single-antenna users transmit simultaneously on the same frequency band and are separated by distinct, not necessarily orthogonal, code sequences. A subspace-based estimation method is proposed to jointly exploit pilots and data to recover channel responses and localization parameters (angles and ranges) of all users. A new kurtosis-based criterion is introduced to accurately estimate the number of propagation paths per user. Unlike conventional techniques, the proposed framework applies seamlessly to near-field, far-field, and hybrid-field conditions. We also define a dynamic, system-aware range boundary that depends on parameters such as pilot length, noise variance, directions of arrival, and targeted localization accuracy. The simulation results confirm that the proposed approach enables precise multiuser channel estimation and localization in challenging multipath environments, without prior knowledge of the users’ field regions.

    2026
    引用
    AI阅读
    加入学术空间
    3FIT-IRSA: Feedback-Integrated Two-phase IRSA with Deep Reinforcement Learning and Its K-Phase Generalization
    Andrei-Valentin Stirbu, Cédric Adjih

    Efficient random access can be used in scenarios with massive numbers of IoT devices. Among modern random access protocols, Irregular Repetition Slotted ALOHA (IRSA) offers excellent asymptotic performance (for large frame sizes), but its finite-frame efficiency is lower and difficult to optimize analytically. In this work, we introduce limited mid-frame feedback to better coordinate users and improve performance: Feedback-Integrated Two-phase IRSA (FIT-IRSA). We formulate IRSA with feedback as a deep reinforcement learning (DRL) problem. Using policy gradient methods, we learn transmission strategies that improve throughput under varying loads, as demonstrated in our simulation results. We further generalize the design to k phases and show that one feedback round captures most of the achievable gain under a fixed slot budget. This provides a practical alternative to classical density-evolution-based optimization, which applies mainly to large frames.

    2026
    引用
    AI阅读
    加入学术空间
    4Extensive Evaluation of R-SCRAW: a Recursive Spectral Clustering Algorithm for Intelligent Grouping in Dense Wireless Networks
    Bruna T. Guedes,Célio Albuquerque,Diego Passos,Fernanda G. O. Passos

    Dense wireless network (DWN) scenarios are more likely to suffer from the problem of hidden terminals and consequently an increase in collisions, as they support a large number of stations in a limited geographic area. While the restricted access window (RAW) mechanism introduced by IEEE 802.11ah reduces the levels of contention by separating nodes into groups with exclusive transmission slots, it does not take hidden terminals into account and, therefore, is still susceptible to collisions caused by them. Recently, a spectral clustering for RAW (SCRAW) algorithm showed potential in finding more efficient RAW groups to reduce hidden terminals, increasing DWN performance. In runtime performance experiments, we observed that SCRAW exhibits a long execution time, which may make its use impractical. This paper is an extension of a previous work that proposed a more efficient SCRAW algorithm, called Recursive SCRAW (R-SCRAW). Our proposal uses a recursive divide and conquer strategy, replacing the full spectral clustering with several executions of a simpler binary split. Simulation results show that the R-SCRAW still significantly outperforms the standard IEEE 802.11ah grouping in terms of collision rate, average packet delay, aggregate network throughput, and number of hidden terminals, while performing just as well as SCRAW but with much lower execution time. Simulations were also performed to compare the behavior of R-SCRAW in scenarios with and without mobility. In scenarios with mobility, the occurrence of hidden terminals increases over time with the distance between stations in each group, decreasing network performance in terms of collision rate and delay compared to simulations without considering mobility. However, even with the performance degradation caused by mobility, R-SCRAW still manages to outperform the standard IEEE 802.11ah grouping under non-mobility conditions. The performance of R-SCRAW was also compared with other grouping algorithms, all under mobility conditions. R-SCRAW presented similar results to K-Means in terms of collision rate, delay, and throughput metrics, and slightly better than the HMR algorithm in terms of throughput, since the initial grouping in HMR is arbitrarily formed.

    2026
    引用
    AI阅读
    加入学术空间
    5Revisit and Distill: Knowledge Distillation for Cybersecurity Language Models
    Hassan N. Noura, Ahmed Hamdi, Yahia Kherza, Guy Pujolle, Ali Chehab

    Pretrained transformer language models such as BERT, RoBERTa, and CodeBERT perform well on code and text classification but are costly to deploy in latency-sensitive settings. We study response-based knowledge distillation (KD) from CodeBERT (124.6M parameters) into BERT-Tiny (4.4M parameters, 28.4× smaller) on two cybersecurity tasks, smart contract vulnerability (SCV) and malicious URL (MURL) classification, using a multi-round protocol in which a frozen teacher is revisited over 10 rounds, the student’s weights are carried forward, and the optimizer and early stopping are reset at each round. Over three seeds, this process narrows the accuracy gap to the teacher from 7.0 to 2.5 points on SCV and from 2.5 to 1.3 points on MURL. To identify the source of these gains, we compare KD and cross-entropy training, each with and without round restarts, under matched training budgets. On SCV, KD improves over cross-entropy in a single continuous run, but restarted cross-entropy training reaches the same accuracy, and combining KD with restarts adds no further gain. On MURL, neither the teacher signal nor restarts improve over longer training on their own, while their combination improves over every partial configuration for all three seeds. Extending the protocol to three teachers, successive multi-teacher distillation reaches 0.893 accuracy on SCV, the highest among the evaluated configurations, at about 2.4× the training epochs of the single-teacher process, and does not improve over single-teacher distillation on MURL. The distilled student requires 141– 188× fewer FLOPs and has 4.2– 7.8× lower single-request GPU latency than the teacher. These results indicate that the benefit of multi-round distillation depends on the dataset and on the interaction between the teacher signal and the training schedule, and that matched-budget controls are needed to attribute it.

    2026
    引用
    AI阅读
    加入学术空间
    立即登录,查看全部 5946 篇论文

    高被引作者

    作者引用发文
    Bernard Fino80912
    Alain Colmerauer11421
    Johannes Huber4273
    Stefan H. Mueller-Weinfurtner4152
    Robert Bauml4152
    Robert Fischer4152
    Sergei Zuyev3572
    François Baccelli3572
    Marc Lebourges3574
    Gérard Battail33227

    高产作者

    作者引用发文
    B. Jecko20645
    Pierre Misme22139
    Guy Pujolle30030
    Lucien Boithias9930
    Odile Macchi6828
    Alain Reineix3628
    Gérard Dubost4927
    Gérard Battail33227
    Pierre Degauque3226
    Michel Joindot9426

    相关期刊

    IEEE Internet of Things Magazine
    IET Networks
    Journal of Network and Systems Management
    Software, Telecommunications and Computer Networks
    Consumer Communications and Networking Conference
    Computer Networks
    Communication Systems and Networks
    International Conference on Information Networking
    Network (Research Triangle Park, N.C.)
    International Symposium on Networks, Computers and Communications