Physical Communication最新文献

Deep learning-enabled integrated sensing and communication via PARAFAC analysis 通过PARAFAC分析实现深度学习集成传感和通信

IF 2.2 4区计算机科学 Q3 ENGINEERING, ELECTRICAL & ELECTRONIC

Physical Communication

Pub Date : 2025-12-08 DOI: 10.1016/j.phycom.2025.102931

Weiwei Jia , Xin Luo , Meifeng Li , Xinman Han , Zhiqiang Yu , Jianhe Du

In this paper, we propose a deep learning-enabled integrated sensing and communication (ISAC) algorithm incorporating PARAFAC analysis to address the challenges faced by ISAC systems. First, a data-driven deep neural network (DNN) based on autoencoders is designed to generate adaptive pilot signals and estimate the channel matrix, ensuring high-precision channel estimation. Second, to enable interference-free multi-user communication, a precoding method is applied to eliminate inter-user interference using the estimated channel state information (CSI) fed back from destination nodes (DNs). Then, DNs construct the received signals from the source node (SN) into a PARAFAC tensor model. A fitting algorithm is used to decompose the tensor for channel estimation and symbol detection, with the estimated CSI serving as the initialization for optimization. Based on the estimated channel, crucial channel parameters such as the angle of arrival (AoA), angle of departure (AoD), and delay are extracted. Furthermore, a low-complexity localization method is employed to determine the positions of the DN and surrounding scatterers using these estimated channel parameters. For the proposed algorithm, we employ the Cramér-Rao Bound (CRB) as a benchmark for evaluation. Simulation results confirm the effectiveness of the proposed algorithm, demonstrating superior performance in both channel estimation accuracy and overall communication quality. Notably, the algorithm maintains excellent ISAC performance even under low compression ratios.

在本文中，我们提出了一种结合PARAFAC分析的深度学习集成传感和通信（ISAC）算法，以解决ISAC系统面临的挑战。首先，设计了基于自编码器的数据驱动深度神经网络（DNN），生成自适应导频信号并估计信道矩阵，保证了信道估计的高精度。其次，为了实现无干扰的多用户通信，采用预编码方法，利用目标节点（DNs）反馈的估计信道状态信息（CSI）消除用户间干扰。然后，DNs将从源节点（SN）接收到的信号构建为PARAFAC张量模型。采用拟合算法对张量进行分解，进行信道估计和符号检测，并将估计的CSI作为初始化进行优化。在估计信道的基础上，提取关键信道参数，如到达角、离开角和时延。此外，采用一种低复杂度的定位方法，利用这些估计的信道参数确定DN和周围散射体的位置。对于所提出的算法，我们采用cram - rao边界（CRB）作为评价的基准。仿真结果验证了该算法的有效性，在信道估计精度和整体通信质量方面均表现出优异的性能。值得注意的是，即使在较低的压缩比下，该算法也保持了良好的ISAC性能。

{"title":"Deep learning-enabled integrated sensing and communication via PARAFAC analysis","authors":"Weiwei Jia , Xin Luo , Meifeng Li , Xinman Han , Zhiqiang Yu , Jianhe Du","doi":"10.1016/j.phycom.2025.102931","DOIUrl":"10.1016/j.phycom.2025.102931","url":null,"abstract":"<div><div>In this paper, we propose a deep learning-enabled integrated sensing and communication (ISAC) algorithm incorporating PARAFAC analysis to address the challenges faced by ISAC systems. First, a data-driven deep neural network (DNN) based on autoencoders is designed to generate adaptive pilot signals and estimate the channel matrix, ensuring high-precision channel estimation. Second, to enable interference-free multi-user communication, a precoding method is applied to eliminate inter-user interference using the estimated channel state information (CSI) fed back from destination nodes (DNs). Then, DNs construct the received signals from the source node (SN) into a PARAFAC tensor model. A fitting algorithm is used to decompose the tensor for channel estimation and symbol detection, with the estimated CSI serving as the initialization for optimization. Based on the estimated channel, crucial channel parameters such as the angle of arrival (AoA), angle of departure (AoD), and delay are extracted. Furthermore, a low-complexity localization method is employed to determine the positions of the DN and surrounding scatterers using these estimated channel parameters. For the proposed algorithm, we employ the Cramér-Rao Bound (CRB) as a benchmark for evaluation. Simulation results confirm the effectiveness of the proposed algorithm, demonstrating superior performance in both channel estimation accuracy and overall communication quality. Notably, the algorithm maintains excellent ISAC performance even under low compression ratios.</div></div>","PeriodicalId":48707,"journal":{"name":"Physical Communication","volume":"74 ","pages":"Article 102931"},"PeriodicalIF":2.2,"publicationDate":"2025-12-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145748483","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Hybrid AI-driven optimization for real-time 6G ad hoc communications using fluid antenna systems 使用流体天线系统的实时6G自组织通信的混合ai驱动优化

IF 2.2 4区计算机科学 Q3 ENGINEERING, ELECTRICAL & ELECTRONIC

Physical Communication

Pub Date : 2025-12-07 DOI: 10.1016/j.phycom.2025.102948

El Miloud Ar-Reyouchi , Ayoub Hadj-Sadek , Kamal Ghoumid , Sami Hage-Ali , Omar Elmazria

Future sixth-generation (6G) Ad Hoc networks must sustain ultra-reliable, low-latency, and energy-efficient connectivity in highly dynamic wireless environments, where interference management, real-time antenna reconfiguration, and computational constraints remain major challenges. Fluid Antenna Systems (FAS) provide additional spatial degrees of freedom through position- and shape-reconfigurable radiating elements, but existing optimization schemes for FAS and next-generation reconfigurable antennas either treat beamforming, phase control, and antenna positioning separately or rely on high-complexity Artificial Intelligence (AI) models that are difficult to deploy under slot-level latency and power budgets. This paper aims to design a unified, low-complexity framework for real-time control of FAS in 6G ad hoc networks. We propose AI-HFASO, a hybrid AI framework in which a Multi-Task Coordination Controller (MTCC) jointly optimizes beamforming, interference mitigation, and antenna positioning by integrating deep learning (DL) for fast beamforming initialization, reinforcement learning (RL) for adaptive element positioning, and a delay-aware genetic algorithm (GA) for phase refinement under latency constraints. The main novelty lies in the joint multi-objective optimization of spectral efficiency, interference suppression, and energy efficiency at the slot level, while reducing computational complexity through lightweight AI modules and a hybrid AI-traditional optimization loop. Simulation results under realistic multi-user, multi-cell 6G scenarios show that AI-HFASO achieves up to 31 % interference reduction, 21 % throughput improvement, and 18 % spectral-efficiency gain, while lowering computational overhead by about 30 % compared to state-of-the-art MIMO, RIS, and AI-based baselines, demonstrating its potential as a scalable and latency-aware solution for FAS-enabled 6G ad hoc networks.

未来的第六代（6G）自组织网络必须在高度动态的无线环境中保持超可靠、低延迟和高能效的连接，在这些环境中，干扰管理、实时天线重构和计算限制仍然是主要挑战。流体天线系统（FAS）通过位置和形状可重构的辐射元件提供了额外的空间自由度，但现有的FAS和下一代可重构天线优化方案要么单独处理波束形成、相位控制和天线定位，要么依赖于高复杂性的人工智能（AI）模型，这些模型很难在槽级延迟和功耗预算下部署。本文旨在为6G自组织网络中FAS的实时控制设计一个统一的、低复杂度的框架。我们提出了AI- hfaso，这是一种混合AI框架，其中多任务协调控制器（MTCC）通过集成用于快速波束形成初始化的深度学习（DL）、用于自适应元件定位的强化学习（RL）和用于延迟约束下相位优化的延迟感知遗传算法（GA），共同优化波束形成、干扰缓解和天线定位。其主要新颖之处在于频谱效率、干扰抑制和槽级能量效率的联合多目标优化，同时通过轻量级AI模块和AI-传统混合优化回路降低计算复杂度。在真实的多用户、多小区6G场景下的仿真结果表明，与最先进的MIMO、RIS和基于ai的基线相比，AI-HFASO实现了高达31%的干扰减少、21%的吞吐量提高和18%的频谱效率增益，同时降低了约30%的计算开销，证明了其作为支持fas的6G自组织网络的可扩展和延迟感知解决方案的潜力。

{"title":"Hybrid AI-driven optimization for real-time 6G ad hoc communications using fluid antenna systems","authors":"El Miloud Ar-Reyouchi , Ayoub Hadj-Sadek , Kamal Ghoumid , Sami Hage-Ali , Omar Elmazria","doi":"10.1016/j.phycom.2025.102948","DOIUrl":"10.1016/j.phycom.2025.102948","url":null,"abstract":"<div><div>Future sixth-generation (6G) Ad Hoc networks must sustain ultra-reliable, low-latency, and energy-efficient connectivity in highly dynamic wireless environments, where interference management, real-time antenna reconfiguration, and computational constraints remain major challenges. Fluid Antenna Systems (FAS) provide additional spatial degrees of freedom through position- and shape-reconfigurable radiating elements, but existing optimization schemes for FAS and next-generation reconfigurable antennas either treat beamforming, phase control, and antenna positioning separately or rely on high-complexity Artificial Intelligence (AI) models that are difficult to deploy under slot-level latency and power budgets. This paper aims to design a unified, low-complexity framework for real-time control of FAS in 6G ad hoc networks. We propose <strong>AI-HFASO</strong>, a hybrid AI framework in which a Multi-Task Coordination Controller (MTCC) jointly optimizes beamforming, interference mitigation, and antenna positioning by integrating deep learning (DL) for fast beamforming initialization, reinforcement learning (RL) for adaptive element positioning, and a delay-aware genetic algorithm (GA) for phase refinement under latency constraints. The main novelty lies in the joint multi-objective optimization of spectral efficiency, interference suppression, and energy efficiency at the slot level, while reducing computational complexity through lightweight AI modules and a hybrid AI-traditional optimization loop. Simulation results under realistic multi-user, multi-cell 6G scenarios show that AI-HFASO achieves up to <strong>31 %</strong> interference reduction, <strong>21 %</strong> throughput improvement, and <strong>18 %</strong> spectral-efficiency gain, while lowering computational overhead by about <strong>30 %</strong> compared to state-of-the-art MIMO, RIS, and AI-based baselines, demonstrating its potential as a scalable and latency-aware solution for FAS-enabled 6G ad hoc networks.</div></div>","PeriodicalId":48707,"journal":{"name":"Physical Communication","volume":"74 ","pages":"Article 102948"},"PeriodicalIF":2.2,"publicationDate":"2025-12-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145748479","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Efficient signal detection in downlink NOMA systems using LSTM-projected layer deep neural networks 基于lstm投影层深度神经网络的下行NOMA系统有效信号检测

IF 2.2 4区计算机科学 Q3 ENGINEERING, ELECTRICAL & ELECTRONIC

Physical Communication

Pub Date : 2025-12-07 DOI: 10.1016/j.phycom.2025.102950

Sebin J. Olickal, Renu Jose

Non-orthogonal multiple access (NOMA) systems help to increase the spectral efficiency of wireless communication and thus support many users simultaneously, making them suitable for next-generation wireless networks. However, signal detection in NOMA systems remains a significant challenge due to inherent interference between users. This paper introduces a novel deep learning (DL) based signal detection method which uses a long-short-term memory projected layer (LSTM-PL), a deep neural network (DNN) model to address this challenge effectively. The proposed approach uses the sequence learning capabilities of LSTMs to capture temporal dependencies in received signals, enabling more accurate detection of superimposed user signals. By incorporating a Projected Layer (PL), the complexity of the detection process is substantially reduced, making it suitable for real-time applications. Extensive simulations demonstrate that the proposed LSTM-PL-based detector achieves a better symbol error rate (SER) compared to traditional signal detection techniques and other DNNs. The method is compared with conventional approaches such as least squares (LS) successive interference cancellation (SIC) and SIC minimum mean square error (MMSE), as well as deep learning methods such as LSTM and gated recurrent unit (GRU). The simulations were conducted using different pilot lengths: 64, 16, 8, and 4. The SER shows that the proposed method leaves behind both conventional and other DNN techniques, offering a robust and efficient solution for signal detection in NOMA systems.

非正交多址（NOMA）系统有助于提高无线通信的频谱效率，从而同时支持多个用户，使其适用于下一代无线网络。然而，由于用户之间的固有干扰，在NOMA系统中信号检测仍然是一个重大挑战。本文介绍了一种新的基于深度学习（DL）的信号检测方法，该方法使用长短期记忆投影层（LSTM-PL），一种深度神经网络（DNN）模型来有效地解决这一挑战。该方法利用lstm的序列学习能力来捕获接收信号中的时间依赖性，从而更准确地检测叠加的用户信号。通过结合投影层（PL），大大降低了检测过程的复杂性，使其适合实时应用。大量的仿真表明，与传统的信号检测技术和其他深度神经网络相比，所提出的基于lstm - pl的检测器具有更好的符号错误率（SER）。将该方法与传统的最小二乘（LS）逐次干扰消除（SIC）和SIC最小均方误差（MMSE）方法以及LSTM和门控循环单元（GRU）等深度学习方法进行了比较。模拟使用不同的飞行员长度：64、16、8和4。SER表明，该方法优于传统的深度神经网络技术和其他深度神经网络技术，为NOMA系统中的信号检测提供了鲁棒和高效的解决方案。

{"title":"Efficient signal detection in downlink NOMA systems using LSTM-projected layer deep neural networks","authors":"Sebin J. Olickal, Renu Jose","doi":"10.1016/j.phycom.2025.102950","DOIUrl":"10.1016/j.phycom.2025.102950","url":null,"abstract":"<div><div>Non-orthogonal multiple access (NOMA) systems help to increase the spectral efficiency of wireless communication and thus support many users simultaneously, making them suitable for next-generation wireless networks. However, signal detection in NOMA systems remains a significant challenge due to inherent interference between users. This paper introduces a novel deep learning (DL) based signal detection method which uses a long-short-term memory projected layer (LSTM-PL), a deep neural network (DNN) model to address this challenge effectively. The proposed approach uses the sequence learning capabilities of LSTMs to capture temporal dependencies in received signals, enabling more accurate detection of superimposed user signals. By incorporating a Projected Layer (PL), the complexity of the detection process is substantially reduced, making it suitable for real-time applications. Extensive simulations demonstrate that the proposed LSTM-PL-based detector achieves a better symbol error rate (SER) compared to traditional signal detection techniques and other DNNs. The method is compared with conventional approaches such as least squares (LS) successive interference cancellation (SIC) and SIC minimum mean square error (MMSE), as well as deep learning methods such as LSTM and gated recurrent unit (GRU). The simulations were conducted using different pilot lengths: 64, 16, 8, and 4. The SER shows that the proposed method leaves behind both conventional and other DNN techniques, offering a robust and efficient solution for signal detection in NOMA systems.</div></div>","PeriodicalId":48707,"journal":{"name":"Physical Communication","volume":"74 ","pages":"Article 102950"},"PeriodicalIF":2.2,"publicationDate":"2025-12-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145747983","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

DSFLS-Net: The multi-band cooperative spectrum sensing based on DSFLS-Net DSFLS-Net：基于DSFLS-Net的多波段协同频谱感知

IF 2.2 4区计算机科学 Q3 ENGINEERING, ELECTRICAL & ELECTRONIC

Physical Communication

Pub Date : 2025-12-06 DOI: 10.1016/j.phycom.2025.102932

Guanghai Xu, Xinran Mao, Yonghua Wang

To enhance detection performance and cross-scenario generalization in dynamic complex environments, an intelligent spectrum sensing method based on dual-scale feature learning is proposed. To fully utilize inter-band correlation information in multi-band signals and adaptively capture underlying patterns in data, this approach combines the Fast Fourier Transform (FFT) and the Discrete Wavelet Transform (DWT) for feature extraction and learning from both the frequency and time-frequency domains. The method first employs an FFT module to extract global frequency-domain features from multi-band signals, obtaining overall energy distribution and spectral characteristics across sub-bands. Simultaneously, a DWT module enables multiresolution time-frequency analysis to mine local time-frequency details. To reduce complexity in broadband detection tasks, a multi-label classification framework is adopted, providing a scalable solution for multi-scenario applications. Furthermore, to address class imbalance in training samples, a focal loss (FL) function is introduced to dynamically adjust learning weights, thereby improving sensing performance in complex environments. Simulations demonstrate that the proposed method achieves excellent detection performance, strong generalization capability, and good robustness across varying SNR conditions and dynamic complex scenarios, offering new insights for multi-band intelligent spectrum sensing.

为了提高动态复杂环境下的检测性能和跨场景泛化能力，提出了一种基于双尺度特征学习的智能频谱感知方法。为了充分利用多波段信号中的带间相关信息并自适应捕获数据中的潜在模式，该方法结合了快速傅里叶变换（FFT）和离散小波变换（DWT），从频域和时频域进行特征提取和学习。该方法首先利用FFT模块从多波段信号中提取全局频域特征，得到各子波段的总体能量分布和频谱特征。同时，DWT模块支持多分辨率时频分析，以挖掘本地时频细节。为了降低宽带检测任务的复杂性，采用了多标签分类框架，为多场景应用提供了可扩展的解决方案。此外，为了解决训练样本中的类不平衡问题，引入焦点损失（focal loss， FL）函数来动态调整学习权值，从而提高复杂环境下的传感性能。仿真结果表明，该方法在不同信噪比条件和动态复杂场景下具有优异的检测性能、较强的泛化能力和较好的鲁棒性，为多频段智能频谱感知提供了新的思路。

{"title":"DSFLS-Net: The multi-band cooperative spectrum sensing based on DSFLS-Net","authors":"Guanghai Xu, Xinran Mao, Yonghua Wang","doi":"10.1016/j.phycom.2025.102932","DOIUrl":"10.1016/j.phycom.2025.102932","url":null,"abstract":"<div><div>To enhance detection performance and cross-scenario generalization in dynamic complex environments, an intelligent spectrum sensing method based on dual-scale feature learning is proposed. To fully utilize inter-band correlation information in multi-band signals and adaptively capture underlying patterns in data, this approach combines the Fast Fourier Transform (FFT) and the Discrete Wavelet Transform (DWT) for feature extraction and learning from both the frequency and time-frequency domains. The method first employs an FFT module to extract global frequency-domain features from multi-band signals, obtaining overall energy distribution and spectral characteristics across sub-bands. Simultaneously, a DWT module enables multiresolution time-frequency analysis to mine local time-frequency details. To reduce complexity in broadband detection tasks, a multi-label classification framework is adopted, providing a scalable solution for multi-scenario applications. Furthermore, to address class imbalance in training samples, a focal loss (FL) function is introduced to dynamically adjust learning weights, thereby improving sensing performance in complex environments. Simulations demonstrate that the proposed method achieves excellent detection performance, strong generalization capability, and good robustness across varying SNR conditions and dynamic complex scenarios, offering new insights for multi-band intelligent spectrum sensing.</div></div>","PeriodicalId":48707,"journal":{"name":"Physical Communication","volume":"74 ","pages":"Article 102932"},"PeriodicalIF":2.2,"publicationDate":"2025-12-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145748484","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Quantum inspired electronic beam scanner with directional jamming mitigation characteristics using time modulation 具有时间调制定向干扰抑制特性的量子激励电子束扫描器

IF 2.2 4区计算机科学 Q3 ENGINEERING, ELECTRICAL & ELECTRONIC

Physical Communication

Pub Date : 2025-12-06 DOI: 10.1016/j.phycom.2025.102953

Avishek Chakraborty , Durbadal Mandal

The proposed work aims to develop an efficient sidelobe cancelling beamformer (SLCB) antenna array for beam scanning and anti-jamming applications in radars. Beam scanning is a technique that directs the antenna radiation patterns across a specific angular region in space to detect a particular object or signal from a specific direction, using either phased arrays or mechanically steered arrays. Time-modulated antenna arrays are alternatives to conventional phased arrays, where the phase of the incoming signals can be adjusted by controlling the ON and OFF times of the array elements. This research addresses the design of an optimized time-modulated beamformer, aiming for a broad scanning coverage of nearly 90˚ in the broadside direction. The scanned beams are generated by proposing quantum algorithm-inspired optimal time schemes, where the ON times of each antenna are optimized to get the optimal radiation characteristics. The beamformer eventually tries to cancel or suppress the sidelobes of the array. The proposed SLCB is designed for an 8-element antenna array, and the incorporation of time-modulation helps to achieve simultaneous beam scanning over a desired region. The scanned beams with reduced sidelobes are further exploited for the strategic null placement in the direction of jammers. The strategic null placement is nothing but the creation of a zero-signal space in the angular region so that the interfering signals can be effectively blocked by focusing on the intended signals. The proposed 8-element timed antenna array is designed to place nulls in a single direction as well as multiple directions for jamming purposes. Further, the adaptive SLCB also aimed to steer the single and multiple nulls in various directions to show the efficiencies, effectiveness, and robustness of the proposed work. The sidelobe-cancelling properties of the beamformer also demonstrate decent performance, suppressing sidelobes below 20 dB for all applications. The radiation efficiency of the proposed electronically controlled beamformer is around 70 %. The proposed SLCB is also compared with other related research works for a comprehensive comparison.

提出的工作旨在开发一种有效的旁瓣对消波束形成器（SLCB）天线阵列，用于雷达的波束扫描和抗干扰应用。波束扫描是一种利用相控阵或机械定向阵，引导天线辐射模式穿过空间中特定角度区域，以探测来自特定方向的特定物体或信号的技术。时间调制天线阵列是传统相控阵的替代方案，其中输入信号的相位可以通过控制阵列元件的开和关时间来调节。本文研究了一种优化的时调制波束形成器的设计，目标是在宽方向上实现近90˚的宽扫描覆盖。通过提出量子算法启发的最优时间方案来产生扫描波束，其中优化每个天线的打开时间以获得最佳的辐射特性。波束形成器最终试图消除或抑制阵列的副瓣。所提出的SLCB是为8元天线阵列设计的，并且时间调制的结合有助于在期望区域内实现同时波束扫描。进一步利用旁瓣减小的扫描波束在干扰器方向上的策略零位置。战略性的零位置只不过是在角区域中创建一个零信号空间，以便通过聚焦预期的信号来有效地阻挡干扰信号。所提出的8元定时天线阵列设计用于在单个方向以及多个方向上放置零点以达到干扰目的。此外，自适应SLCB还旨在在不同方向上引导单个和多个null，以显示所提出工作的效率，有效性和鲁棒性。波束形成器的副瓣消除特性也表现出良好的性能，在所有应用中都能抑制20 dB以下的副瓣。该电子束形成器的辐射效率约为70%。并与其他相关研究成果进行了比较。

{"title":"Quantum inspired electronic beam scanner with directional jamming mitigation characteristics using time modulation","authors":"Avishek Chakraborty , Durbadal Mandal","doi":"10.1016/j.phycom.2025.102953","DOIUrl":"10.1016/j.phycom.2025.102953","url":null,"abstract":"<div><div>The proposed work aims to develop an efficient sidelobe cancelling beamformer (SLCB) antenna array for beam scanning and anti-jamming applications in radars. Beam scanning is a technique that directs the antenna radiation patterns across a specific angular region in space to detect a particular object or signal from a specific direction, using either phased arrays or mechanically steered arrays. Time-modulated antenna arrays are alternatives to conventional phased arrays, where the phase of the incoming signals can be adjusted by controlling the ON and OFF times of the array elements. This research addresses the design of an optimized time-modulated beamformer, aiming for a broad scanning coverage of nearly 90˚ in the broadside direction. The scanned beams are generated by proposing quantum algorithm-inspired optimal time schemes, where the ON times of each antenna are optimized to get the optimal radiation characteristics. The beamformer eventually tries to cancel or suppress the sidelobes of the array. The proposed SLCB is designed for an 8-element antenna array, and the incorporation of time-modulation helps to achieve simultaneous beam scanning over a desired region. The scanned beams with reduced sidelobes are further exploited for the strategic null placement in the direction of jammers. The strategic null placement is nothing but the creation of a zero-signal space in the angular region so that the interfering signals can be effectively blocked by focusing on the intended signals. The proposed 8-element timed antenna array is designed to place nulls in a single direction as well as multiple directions for jamming purposes. Further, the adaptive SLCB also aimed to steer the single and multiple nulls in various directions to show the efficiencies, effectiveness, and robustness of the proposed work. The sidelobe-cancelling properties of the beamformer also demonstrate decent performance, suppressing sidelobes below 20 dB for all applications. The radiation efficiency of the proposed electronically controlled beamformer is around 70 %. The proposed SLCB is also compared with other related research works for a comprehensive comparison.</div></div>","PeriodicalId":48707,"journal":{"name":"Physical Communication","volume":"74 ","pages":"Article 102953"},"PeriodicalIF":2.2,"publicationDate":"2025-12-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145748478","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Efficient convolutional LDPC decoding for Nakagami-m fading channels in 5G-NR 5G-NR中agami-m衰落信道的高效卷积LDPC解码

IF 2.2 4区计算机科学 Q3 ENGINEERING, ELECTRICAL & ELECTRONIC

Physical Communication

Pub Date : 2025-12-06 DOI: 10.1016/j.phycom.2025.102952

Nour El Houda Zareb , Mohamed Azni , Meriem Gagaoua , Reda Kasmi

The increasing complexity of 5G New Radio (5G-NR) systems has reinforced the need for efficient and robust low-density parity-check (LDPC) decoders capable of meeting stringent requirements in terms of reliability, latency, and throughput. Among the adopted solutions, protograph-based quasi-cyclic (PB-QC) LDPC codes offer an attractive balance between performance and hardware efficiency. Conventional belief propagation (BP) decoding performs well under additive white gaussian noise (AWGN) but its performance degrades under realistic fading conditions, limiting its applicability in diverse environments. Although BP neural networks have shown potential, most are tightly tied to the Tanner graph or restricted to simple noise models, leaving a gap in flexible, generalizable decoding strategies.

In this work, we present an efficient fully neural LDPC decoder based on convolutional neural networks (CNNs), designed for PB-QC codes with a rate of

R = 0.5

and trained under various Nakagami-m fading conditions. The proposed CNN decoder improves adaptability across varying Nakagami-m fading scenarios without requiring architectural changes, while its lightweight structure significantly reduces computational complexity, achieving more than 67 % lower inference time compared to belief propagation neural network (BP-NN) baselines. A signal-repetition preprocessing technique further enhances performance in severe fading regimes. Simulation results show that our CNN decoder achieves competitive bit error rate (BER) performance compared to BP-NN baselines. These results position CNN-based LDPC decoding as a viable and efficient solution for next-generation wireless systems operating under dynamic and challenging channel conditions.

5G新无线电（5G- nr）系统的复杂性日益增加，加强了对高效、稳健的低密度奇偶校验（LDPC）解码器的需求，这些解码器能够满足可靠性、延迟和吞吐量方面的严格要求。在采用的解决方案中，基于原型的准循环（PB-QC） LDPC码在性能和硬件效率之间提供了一个有吸引力的平衡。传统的信念传播（BP）译码在加性高斯白噪声（AWGN）下性能良好，但在实际衰落条件下性能下降，限制了其在多种环境下的适用性。尽管BP神经网络显示出了潜力，但大多数都与坦纳图紧密相连，或者局限于简单的噪声模型，在灵活、通用的解码策略上留下了空白。在这项工作中，我们提出了一个基于卷积神经网络（cnn）的高效全神经LDPC解码器，设计用于率为R=0.5的PB-QC码，并在各种Nakagami-m衰落条件下进行训练。所提出的CNN解码器在不需要改变架构的情况下提高了不同Nakagami-m衰落场景的适应性，同时其轻量级结构显著降低了计算复杂度，与信念传播神经网络（BP-NN）基线相比，推理时间降低了67%以上。信号重复预处理技术进一步提高了在严重衰落情况下的性能。仿真结果表明，与BP-NN基线相比，我们的CNN解码器实现了具有竞争力的误码率（BER）性能。这些结果使基于cnn的LDPC解码成为在动态和具有挑战性的信道条件下运行的下一代无线系统的可行且有效的解决方案。

{"title":"Efficient convolutional LDPC decoding for Nakagami-m fading channels in 5G-NR","authors":"Nour El Houda Zareb , Mohamed Azni , Meriem Gagaoua , Reda Kasmi","doi":"10.1016/j.phycom.2025.102952","DOIUrl":"10.1016/j.phycom.2025.102952","url":null,"abstract":"<div><div>The increasing complexity of 5G New Radio (5G-NR) systems has reinforced the need for efficient and robust low-density parity-check (LDPC) decoders capable of meeting stringent requirements in terms of reliability, latency, and throughput. Among the adopted solutions, protograph-based quasi-cyclic (PB-QC) LDPC codes offer an attractive balance between performance and hardware efficiency. Conventional belief propagation (BP) decoding performs well under additive white gaussian noise (AWGN) but its performance degrades under realistic fading conditions, limiting its applicability in diverse environments. Although BP neural networks have shown potential, most are tightly tied to the Tanner graph or restricted to simple noise models, leaving a gap in flexible, generalizable decoding strategies.</div><div>In this work, we present an efficient fully neural LDPC decoder based on convolutional neural networks (CNNs), designed for PB-QC codes with a rate of <span><math><mrow><mi>R</mi><mo>=</mo><mn>0.5</mn></mrow></math></span> and trained under various Nakagami-<em>m</em> fading conditions. The proposed CNN decoder improves adaptability across varying Nakagami-<em>m</em> fading scenarios without requiring architectural changes, while its lightweight structure significantly reduces computational complexity, achieving more than 67 % lower inference time compared to belief propagation neural network (BP-NN) baselines. A signal-repetition preprocessing technique further enhances performance in severe fading regimes. Simulation results show that our CNN decoder achieves competitive bit error rate (BER) performance compared to BP-NN baselines. These results position CNN-based LDPC decoding as a viable and efficient solution for next-generation wireless systems operating under dynamic and challenging channel conditions.</div></div>","PeriodicalId":48707,"journal":{"name":"Physical Communication","volume":"74 ","pages":"Article 102952"},"PeriodicalIF":2.2,"publicationDate":"2025-12-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145748482","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Joint UAV flight position and phase shift matrix design for active RIS-assisted communication systems 主动ris辅助通信系统的联合无人机飞行位置和相移矩阵设计

IF 2.2 4区计算机科学 Q3 ENGINEERING, ELECTRICAL & ELECTRONIC

Physical Communication

Pub Date : 2025-12-05 DOI: 10.1016/j.phycom.2025.102941

Mengge Shen , Yilong Liu , Dongxing Li , Jun Zhang

The active reconfigurable intelligent surface (RIS) is able to actively reflect signals with amplification and overcome the multiplicative fading effect. Therefore, introducing the active RIS into the unmanned aerial vehicle (UAV) communication systems can further enhance the communication performance. In this paper, we investigate an active RIS-assisted UAV communication system in which the UAV employed a single directional antenna is functioned as an airborne BS. Our objective is to maximize the achievable rate with the power constraint at the active RIS by jointly designing the three-dimensional (3D) flight position for the UAV and the phase shift matrix at the active RIS. Firstly, the phase shift matrix at the active RIS is derived by using the phase alignment approach, and the 3D coordinates for the UAV are derived based on the fixed point iteration method. Then, an alternating optimization algorithm is introduced to obtain the aforementioned variables. Finally, the numerical results are conducted to verify the effectiveness of the proposed method.

主动可重构智能曲面（RIS）能够主动反射放大信号，克服乘衰落效应。因此，在无人机通信系统中引入主动RIS可以进一步提高通信性能。在本文中，我们研究了一种主动ris辅助无人机通信系统，其中无人机采用单向天线作为机载BS。我们的目标是通过联合设计无人机的三维（3D）飞行位置和主动RIS处的相移矩阵，在功率约束下最大化可实现速率。首先，采用相位对准法推导了主动RIS处的相移矩阵，并基于不动点迭代法推导了无人机的三维坐标；然后，引入交替优化算法来获取上述变量。最后通过数值算例验证了所提方法的有效性。

引用次数: 0

Data augmentation via noise injection and VMD for deep learning-based modulation recognition in few-shot scenario 基于噪声注入和VMD的数据增强，用于基于深度学习的调制识别

IF 2.2 4区计算机科学 Q3 ENGINEERING, ELECTRICAL & ELECTRONIC

Physical Communication

Pub Date : 2025-12-03 DOI: 10.1016/j.phycom.2025.102945

Tao Chen , Shilian Zheng

Deep learning has been widely adopted in radio modulation recognition. However, when training data is insufficient, deep learning models are prone to overfitting, leading to a lack of generalization ability. In this letter, we use noise injection as a data augmentation strategy and combine it with variational mode decomposition (VMD) to suppress irrelevant noise components that maintain structural fidelity while exhibiting perturbation diversity. This modulation recognition framework is referred to as VMD-ResNet. We conduct experiments by adding additive white Gaussian noise (AWGN), additive general Gaussian noise (AGGN), and pink noise. Compared to the IQ-ResNet method without augmentation, the findings indicate that the level of improvement varies among different modulation types depending on the specific noise added. Comparative experiments confirm that the proposed VMD-ResNet achieves higher overall recognition performance than other few-shot methods. Additionally, we assess the model using unseen noise sequences to further validate its generalization capability. The experimental results demonstrate the method’s enhanced effectiveness for unseen noise.

深度学习在无线电调制识别中得到了广泛的应用。然而，当训练数据不足时，深度学习模型容易出现过拟合，导致泛化能力不足。在这封信中，我们使用噪声注入作为数据增强策略，并将其与变分模态分解（VMD）相结合，以抑制不相关的噪声成分，在保持结构保真度的同时表现出扰动多样性。这个调制识别框架被称为VMD-ResNet。我们通过加性高斯白噪声（AWGN）、加性一般高斯噪声（AGGN）和粉红噪声进行实验。与没有增强的IQ-ResNet方法相比，研究结果表明，不同调制类型的改进水平取决于添加的特定噪声。对比实验表明，所提出的VMD-ResNet方法的整体识别性能优于其他的少弹方法。此外，我们使用看不见的噪声序列来评估模型，以进一步验证其泛化能力。实验结果表明，该方法对不可见噪声有较好的抑制效果。

{"title":"Data augmentation via noise injection and VMD for deep learning-based modulation recognition in few-shot scenario","authors":"Tao Chen , Shilian Zheng","doi":"10.1016/j.phycom.2025.102945","DOIUrl":"10.1016/j.phycom.2025.102945","url":null,"abstract":"<div><div>Deep learning has been widely adopted in radio modulation recognition. However, when training data is insufficient, deep learning models are prone to overfitting, leading to a lack of generalization ability. In this letter, we use noise injection as a data augmentation strategy and combine it with variational mode decomposition (VMD) to suppress irrelevant noise components that maintain structural fidelity while exhibiting perturbation diversity. This modulation recognition framework is referred to as VMD-ResNet. We conduct experiments by adding additive white Gaussian noise (AWGN), additive general Gaussian noise (AGGN), and pink noise. Compared to the IQ-ResNet method without augmentation, the findings indicate that the level of improvement varies among different modulation types depending on the specific noise added. Comparative experiments confirm that the proposed VMD-ResNet achieves higher overall recognition performance than other few-shot methods. Additionally, we assess the model using unseen noise sequences to further validate its generalization capability. The experimental results demonstrate the method’s enhanced effectiveness for unseen noise.</div></div>","PeriodicalId":48707,"journal":{"name":"Physical Communication","volume":"74 ","pages":"Article 102945"},"PeriodicalIF":2.2,"publicationDate":"2025-12-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145748480","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A high data rate ambient backscatter DCSK communication system based on joint permutation index and code index modulations 基于联合排列索引和码索引调制的高数据速率环境后向散射DCSK通信系统

IF 2.2 4区计算机科学 Q3 ENGINEERING, ELECTRICAL & ELECTRONIC

Physical Communication

Pub Date : 2025-12-01 DOI: 10.1016/j.phycom.2025.102906

Doaa S. Ibrahim, Fadhil S. Hasan

In this paper, a joint permutation index and code index modulations differential chaos shift keying-based ambient backscatter communication (JPCIM-DCSK-AmBC) is proposed to improve the system performance and data rate. The JPCIM-DCSK-AmBC signal in the proposed system is transmitted using a backscatter device, with multiple bits encoded in a radio frequency (RF) source symbol via joint permutation index and a modulation technique based on code indexing. The information bits are split into two blocks of

m_{p}

and

m_{c}

bits, where

m_{p}

bits are utilized for permutation index and

m_{c}

bits are utilized for code index bits. The analytic bit error rate (BER) of the JPCIM-DCSK-AmBC over a flat fading channel is derived and verified with the simulation results. The experimental results demonstrate that the suggested system has lower error rate and higher throughput than conventional methods.

为了提高系统性能和数据速率，提出了一种组合排列索引和编码索引调制的基于差分混沌移位键控的环境后向散射通信（JPCIM-DCSK-AmBC）。该系统中的JPCIM-DCSK-AmBC信号采用后向散射装置，通过联合排列索引和基于码索引的调制技术在射频（RF）源符号中编码多个比特。信息位被分成mp和mcbits两个块，其中mp bits用于排列索引，mcbits用于代码索引位。推导了JPCIM-DCSK-AmBC在平坦衰落信道上的分析误码率，并用仿真结果进行了验证。实验结果表明，该系统具有较低的错误率和较高的吞吐量。

{"title":"A high data rate ambient backscatter DCSK communication system based on joint permutation index and code index modulations","authors":"Doaa S. Ibrahim, Fadhil S. Hasan","doi":"10.1016/j.phycom.2025.102906","DOIUrl":"10.1016/j.phycom.2025.102906","url":null,"abstract":"<div><div>In this paper, a joint permutation index and code index modulations differential chaos shift keying-based ambient backscatter communication (JPCIM-DCSK-AmBC) is proposed to improve the system performance and data rate. The JPCIM-DCSK-AmBC signal in the proposed system is transmitted using a backscatter device, with multiple bits encoded in a radio frequency (RF) source symbol via joint permutation index and a modulation technique based on code indexing. The information bits are split into two blocks of <span><math><msub><mi>m</mi><mi>p</mi></msub></math></span> and <span><math><mrow><msub><mi>m</mi><mi>c</mi></msub><mspace></mspace></mrow></math></span>bits, where <span><math><msub><mi>m</mi><mi>p</mi></msub></math></span> bits are utilized for permutation index and <span><math><msub><mi>m</mi><mi>c</mi></msub></math></span> bits are utilized for code index bits. The analytic bit error rate (BER) of the JPCIM-DCSK-AmBC over a flat fading channel is derived and verified with the simulation results. The experimental results demonstrate that the suggested system has lower error rate and higher throughput than conventional methods.</div></div>","PeriodicalId":48707,"journal":{"name":"Physical Communication","volume":"73 ","pages":"Article 102906"},"PeriodicalIF":2.2,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145623675","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Hybrid LSTM-actor-critic framework for temporal-spatial wideband beam tracking in 6G THz massive MIMO 6G THz大规模MIMO中时空宽带波束跟踪的混合LSTM-actor-critic框架

IF 2.2 4区计算机科学 Q3 ENGINEERING, ELECTRICAL & ELECTRONIC

Physical Communication

Pub Date : 2025-11-30 DOI: 10.1016/j.phycom.2025.102933

Siva Kumar T, Jeyakumar P

Efficient beam tracking is essential for stable high-throughput communication in ultra-wideband terahertz (THz) massive MIMO systems. However, existing reinforcement learning (RL)-based methods often overlook temporal channel variations and beam squint effects inherent to wideband THz links. To address this, we propose a Hybrid Long Short-Term Memory–Actor-Critic (LSTM–A2C) framework that integrates temporal prediction and spatial decision-making for adaptive beam tracking in 6G THz systems. The LSTM captures beam dynamics over time, while the Actor–Critic learner adjusts beamforming vectors in real time under mobility and hardware impairments. Simulations under realistic THz conditions (with phase noise, ADC quantization, and beam squint) show that LSTM–A2C achieves 15–25 % higher tracking accuracy and 30–40 % lower error compared to DQN, PPO, and EKF methods. It also delivers 3.5 bps/Hz spectral efficiency at 10 dB SNR, 2.1 bits/Joule energy efficiency, and a fairness index of 0.99 in multi-user settings. The reward curve shows faster convergence ( ≈  200 epochs), validating the proposed framework’s efficiency.

在超宽带太赫兹（THz）大规模MIMO系统中，高效的波束跟踪是稳定高吞吐量通信的关键。然而，现有的基于强化学习（RL）的方法往往忽略了宽带太赫兹链路固有的时间信道变化和波束斜视效应。为了解决这个问题，我们提出了一个混合长短期记忆-行为者-批评（LSTM-A2C）框架，该框架集成了6G太赫兹系统中自适应波束跟踪的时间预测和空间决策。LSTM捕获随时间变化的波束动态，而Actor-Critic学习器在移动和硬件损坏情况下实时调整波束形成矢量。在现实太赫兹条件下（含相位噪声、ADC量化和波束斜视）的仿真表明，与DQN、PPO和EKF方法相比，LSTM-A2C的跟踪精度提高15 - 25%，误差降低30 - 40%。在10db信噪比下，它还提供3.5 bps/Hz的频谱效率，2.1比特/焦耳的能量效率，多用户设置下的公平性指数为0.99。奖励曲线显示出更快的收敛速度（ ≈ 200 epoch），验证了所提出框架的效率。

{"title":"Hybrid LSTM-actor-critic framework for temporal-spatial wideband beam tracking in 6G THz massive MIMO","authors":"Siva Kumar T, Jeyakumar P","doi":"10.1016/j.phycom.2025.102933","DOIUrl":"10.1016/j.phycom.2025.102933","url":null,"abstract":"<div><div>Efficient beam tracking is essential for stable high-throughput communication in ultra-wideband terahertz (THz) massive MIMO systems. However, existing reinforcement learning (RL)-based methods often overlook temporal channel variations and beam squint effects inherent to wideband THz links. To address this, we propose a Hybrid Long Short-Term Memory–Actor-Critic (LSTM–A2C) framework that integrates temporal prediction and spatial decision-making for adaptive beam tracking in 6G THz systems. The LSTM captures beam dynamics over time, while the Actor–Critic learner adjusts beamforming vectors in real time under mobility and hardware impairments. Simulations under realistic THz conditions (with phase noise, ADC quantization, and beam squint) show that LSTM–A2C achieves 15–25 % higher tracking accuracy and 30–40 % lower error compared to DQN, PPO, and EKF methods. It also delivers 3.5 bps/Hz spectral efficiency at 10 dB SNR, 2.1 bits/Joule energy efficiency, and a fairness index of 0.99 in multi-user settings. The reward curve shows faster convergence ( ≈  200 epochs), validating the proposed framework’s efficiency.</div></div>","PeriodicalId":48707,"journal":{"name":"Physical Communication","volume":"74 ","pages":"Article 102933"},"PeriodicalIF":2.2,"publicationDate":"2025-11-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145748481","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0