With the rapid evolution of smart manufacturing, achieving high-precision and real-time tool condition monitoring has become increasingly crucial for maintaining production efficiency and system reliability. To address this challenge, this work proposes a tool wear fault diagnosis framework that integrates deep learning with collaborative edge computing. At the core of this framework, a physically motivated and hierarchically coupled spatial-temporal modeling network, namely Improved Multiscale Networks (IMSNet), is developed. Unlike conventional CNN-LSTM-attention architectures, IMSNet is designed according to the intrinsic degradation characteristics of tool wear vibration signals. Specifically, a structured multiscale convolutional module is employed to extract spatial representations that capture heterogeneous frequency-axis interactions, while an LSTM network simultaneously models the temporal evolution embedded in raw vibration signals to characterize cumulative degradation dynamics. The complementary spatial and temporal features are then fused through a multi-head attention mechanism for adaptive reliability-aware feature reweighting, enabling robust representation learning under non-stationary machining conditions. To meet stringent real-time requirements in Industrial IoT systems, we further design a cloud-edge-device collaboration (CEDC) framework for adaptive task offloading and low-latency inference. The framework decomposes diagnostic workloads across device, edge, and cloud layers and dynamically coordinates computation within scheduling windows, thereby improving system responsiveness. Experiments demonstrate that IMSNet achieves 97.81% accuracy on the self-collected industrial dataset and 98.93% on the public PHM benchmark, and showing stable and generalizable performance under varying operating conditions. Meanwhile, the CEDC framework reduces task off-loading latency by up to 17.4% compared with the advanced offloading strategies, while maintaining real-time responsiveness under varying workload conditions. To facilitate reproducibility and further research, the source code is publicly available at: https://github.com/lidongyang1/tool-fault-diagnosis-IIoT.

Synthetic aperture radar (SAR) is widely used for water body detection due to its efficiency and ability to operate in all weather conditions. However, its scattering properties and single-polarization limitations pose challenges for data extraction and reduce the accuracy of water body detection algorithms. To mitigate this limitation, recent studies have focused on transforming SAR datasets into electro-optical (EO) image modalities through cross-modal translation models, aiming to enhance multispectral feature interpretability. However, such transformation frameworks require substantial computational power, which compromises the real-time processing capabilities critical for rapid disaster response, such as a flood. In this letter, we propose a lightweight SAR water body detection framework that integrates knowledge distillation and channel attention. A teacher network trained on rich EO data guides an SAR-specific student model, with both employing attention branches. The student’s attention is supervised by the teacher to enhance SAR feature extraction via attention-aligned distillation. Evaluated on the Sen1Floods11 benchmark dataset, our experimental results outperform the baseline model by 3.5% in intersection over union (IoU).

With the continuous development of 6G technology and the Internet of Things, small terminal devices are gradually joining deep model training through wireless networks, leading to the evolution of federated learning. In comparison to traditional centralized learning, federated learning not only leverages the computational power of individual terminals but also ensures the security of terminal data. However, the increasing number of devices poses new requirements on resource utilization in federated learning at scale. In this paper, we aim to address these challenges by proposing an energy-efficient and adaptive resource allocation strategy for wireless heterogeneous layered federated learning model (HLFLM). Specifically, we deploy both macro base stations and multiple micro base stations to construct a HLFLM, and perform resource allocation for subcarriers and power optimization. This approach focuses on optimizing energy consumption in federated learning networks while enhancing scalability and real-time performance of wireless communication. Experimental results demonstrate the effectiveness of the proposed method in medium-sized scenarios.

Deep reinforcement learning models have achieved promising results in the path planning problem for uncrewed aerial vehicles (UAVs). However, their update mechanisms can lead to overestimation and poor stability. This study addresses these issues by using a more realistic reward function, assigning priorities to experiences in the experience replay pool, and employing double decoupling of state and action values in Q-networks. We train the improved Deep Q-Network (DQN) algorithm for three-dimensional environment simulation experiments in a simulated environment. The 3D simulation experiments compare the algorithm with the A-star and unimproved DQN algorithms. The experimental results show that the algorithm has been improved, demonstrating its enhanced performance in the final path planning results. Moreover, the final testing results reveal that the UAV can safely reach the target point from the starting point.

With the development of Internet of Medical Things(IOMT), a lot of medical data are stored and released for both scientific research and practical applications. Accurate medical data is very valuable, but it also brings a huge risk of privacy leakage. Moreover, improving the privacy of data often leads to the reduction of data validity. Privacy and effectiveness are in conflict, and their balance is a typical multi-objective optimization problem (MOP). In this paper, we try to use differential privacy to disturb medical data to protect personal privacy. We propose the Environment Switching Algorithm (ESA) based on evolutionary learning to solve this MOP. ESA has excellent performance, which can ensure convergence speed and optimization performance at the same time. The result of optimization is a pareto front (PF) of huge scale, which includes solutions with different characteristics. We put forward a method of double clustering to select the appropriate solution from PF. Based on the above, we conclude the whole method as Flexible Differential Privacy Algorithm based on Evolutionary Learning (FDPEL). FDPEL can realize flexible differential privacy for medical data, while ensuring data privacy and data validity. FDPEL is suitable for privacy protection of medical data of different scales, which makes it have a practical applications value.

Automatic modulation recognition (AMR) of underwater acoustic (UWA) signals is incredibly challenging due to the complexity of UWA channels and the severity of ocean noise. In the presence of noise interference, single-modal features fail to fully represent the characteristics of different modulated signals. While the in-phase/quadrature (I/Q) and time-frequency maps can adequately represent the signal features in the time, frequency, and time-frequency domains, the direct integration of the two modalities is ineffective because of the variations in shape, information granularity, and noise manifestation. To address the low recognition rate caused by the above issues, we propose a two-stream transformer (TSTR) based network for AMR of UWA signals. First, the input pre-processing layer obtains the I/Q and time-frequency features from the received signals. Then, the feature capture layer extracts high-dimensional signal features in the time, frequency, and time-frequency domains. Finally, the classification layer estimates the modulation of the signals. A multi-head self-attention module with adaptive soft thresholding is used in the feature capture layer to provide noise reduction and redundant feature rejection while retaining context information. Moreover, multi-scale ghost convolution is employed to address the inability of the transformer to efficiently extract spatial characteristics from the signals. Results are presented using real UWA channels from the Watermark dataset for two different seas which show that the TSTR improves recognition by 1.2% and 5.9% over the best existing model. Further, it has better generalization capabilities and the model has a small number of parameters so the time complexity is low.

With the rapid development of software-defined networking/network function virtualization (NFV) technologies, service function chaining (SFC) has become a key enabler for end-to-end service provisioning in future networks. In the Internet of Things (IoT), the highly dynamic nature of the network environment demands flexible and adaptive mechanisms for dynamic SFC deployment to fully utilize network resources while meeting the service requirements. Although reinforcement learning (RL) techniques offer a promising approach to dynamic SFC deployment, the learning delay of RL may limit its prompt response to sudden changes in network state and/or service demand. To address this challenge in this article, we propose to employ a deep$Q$-learning network (DQN) method for dynamic SFC deployment combined with a tidal virtual machine (TVM) control mechanism for adaptive virtual machine (VM) auto-scaling. We present a tidal DQN framework (TDQNF) that integrates the DQN method and TVM control in the ETSI NFV architecture and develop the algorithms for implementing DQN-based decisions for SFC deployment and TVM control for VM scaling. The performance of the TDQNF framework with the proposed algorithms has been evaluated through extensive simulation experiments. The obtained experimental results verify the effectiveness of the proposed scheme and indicate better performance in terms of system delay, packet loss, and load balancing in large-scale networks compared to existing methods.

Against the backdrop of the increasingly scarce spectrum, the integrated sensing and communication (ISAC) is emerging as a promising technology to enable simultaneous sensing and communication between vehicles and their surroundings in the vehicular networks. However, to realize ISAC, an effective beamforming design is essential. Motivated by this, this work investigates the beamforming scheme for ISAC-based vehicular networks and a Transformer-based predictive beamforming approach is proposed. In details, a novel transmission protocol is developed to bypass the need for the roadside unit (RSU) to acquire channel state information (CSI) and historical channel parameters. It can establishe a direct relationship between the signal echoes and beamforming matrix in data-driven manner to reduce the signaling overhead. Then, an optimization problem maximizing the communication sum rate is formulated to design the optimal beamforming scheme. Due to the nonconvex nature of the objective function and constraints, a penalty-based method is exploited to transform it to an unconstrained optimization problem and the Echoes-based Convolution Transformer Network (ECT-Net) is proposed to solve it. As a realization of the ECT-Net, the convolutional module and attention mechanism are integrated to jointly capture the reflected echoes’ local and global spatial dependencies, improving the performance of predictive beamforming. Extensive simulations are carried out and show that the proposed Transformer-based predictive beamforming method can achieve higher communication sum rate than state-of-the-art beamforming methods.

In this paper, we study the problem of mobile user pairing in non-orthogonal multi-access (NOMA) backscatter communication networks, in which re-pairing users after a change in the communication link can bring significant computational overhead. Firstly, We propose a user pairing scheme that employs the Kuhn-Munkres (KM) algorithm for the uplink of the monostatic backscatter communication network, which can effectively avoid the problem of frequent sorting mobile users. Secondly, to address the NOMA principle violations problem (NPVP) caused by mobile users, we propose a modified role switching (RS) technique based on the difference of user channel gain in which paired users switch their roles based on the magnitude of their channel gain after moving. Finally, we propose a dynamic power allocation scheme for mobile users to maximize the system throughput after pairing. In backscatter communication, the difference in power levels between paired users is achieved by setting different reflection coefficients. At the same time, it is necessary to consider the balance between data transmission and energy harvesting, and the reflection coefficient must not exceed the predetermined upper limit. Simulation results indicate that our proposed user pairing scheme outperforms conventional schemes in terms of the number of decoded users and system throughput.

Ship detection in synthetic aperture radar (SAR) images using a convolutional neural network (CNN) has become a hotspot. However, dense multi-size ship targets in inshore scenes, proximity of ship targets to man-made targets, and datasets containing only one class limit the performance of ship detection methods. In pursuit of attaining heightened performance, most existing methods endeavor to augment the number of parameters and enhance network complexity. To address these problems, this letter proposes an anchor-free ghost feature extraction and cross-scale interaction network (GFECSI-Net), which improves detection performance through an efficient implementation, thereby avoiding the increase in the number of parameters or network complexity. First, to enhance the capability of feature extraction for ship targets, a multi-scale adaptive feature pyramid network (MSAFPN) is proposed to realize intensive information interaction and cross-scale feature fusion between different feature maps. Meanwhile, a selective efficient channel attention module (SECAM) is designed to enable the network to prioritize channels that better characterize ship targets. Besides, a GPU-efficient backbone for generating ghost feature maps and a task alignment detection head are integrated into GFECSI-Net. Comparison results with nine state-of-the-art CNN methods on a high-resolution SAR image dataset (HRSID) and a large-scale multi-class SAR target dataset (MSAR-1.0) indicate that GFECSI-Net achieves superior detection performance in F1-Score and mAP with a small number of parameters.

Network Virtualization (NV) is an emerging network dynamic planning technique to overcome network rigidity. As its necessary challenge, Virtual Network Embedding (VNE) enhances the scalability and flexibility of the network by decoupling the resources and services of the underlying physical network. For future multi-domain physical network modeling with the characteristics of dynamics, heterogeneity, privacy, and real-time, the existing related works perform unsatisfactorily. Federated learning (FL) jointly optimizes the network by sharing parameters among multiple parties and is widely used to address data privacy and data silos. Aiming at the NV challenge of multi-domain physical networks, this work is the first to propose using FL to model VNE, and presents a VNE architecture based on Horizontal Federated Learning (HFL) (HFL-VNE). Specifically, combined with the distributed training paradigm of FL, we deploy local servers in each physical domain, which can effectively focus on local features and reduce resource fragmentation. A global server is deployed to aggregate and share training parameters, which enhances local data privacy and significantly improves learning efficiency. Furthermore, we deploy the Deep Reinforcement Learning (DRL) model in each server to dynamically adjust and optimize the resource allocation of the multi-domain physical network. In DRL-assisted FL, HFL-VNE jointly optimizes decision-making through specific local and federated reward mechanisms and loss functions. Finally, the superiority of HFL-VNE is proved by combining simulation experiments and comparing it with related works.

Millimeter wave radar-based human activity recognition (HAR) technology has received much attention as a research hot-spot in recent years. Previous researches have demonstrated the feasibility of using millimeter wave radar for HAR. While existing work has achieved excellent performance in ideal environments, its application in life is still limited due to the intensive data collection required, the additional training needed to adapt to new domains (i.e., environments, people, and locations), and the high computational complexity associated with voxelization. To solve the above problems, we propose the radar point cloud recognition system RPCRS, which is capable of accurately recognizing human activities from noisy environments, has promising recognition performance for new users, environments and locations, and significantly reduces the computational overhead during system training. Firstly, RPCRS use the velocity information of the clustered point cloud data to extract the human activity subjects from the noisy background. Then, the size of the extracted non-uniform point cloud data is unified by removing or adding the number of point clouds. Secondly, in order to enhance the robustness of the system and reduce the data collection effort, we designed a data enhancement framework based on correlation between point cloud data and human activity changes. Finally, a lightweight neural network based on a multilayer perceptron (MLP) is used to classify the raw point cloud data of human activities, which reduces the computational complexity and memory requirements associated with voxelization. We evaluate our system with 5 different activities, which attains average accuracy of 95.40%. In addition, we evaluate the performance of the system in a new environment and with new users, which obtains an average accuracy of 94.53% and 95.08%, respectively.

The Transformer is one of the mainstream methods in computer vision. Most Transformer based architectures focus on the design of spatial attention and optimizing the computational complexity of high resolution of pixels in image but pay little attention to modeling channel dependencies and optimizing the computational complexity associated with a large number of channels. In this paper, we propose a new channel window based self-attention mechanism and apply two consecutive transformer layers to capture global channel dependencies through permuting channel layer, which can greatly reduce the computational complexity caused by a large number of channels. Meanwhile, a new linear layer for channel attention is proposed, which eliminates the need for position bias in Transformer. The proposed method can be conveniently appended to the existing image classification architectures in parallel with minimal modification. We demonstrate the feasibility of the proposed method on the state-of-the-art transformer-based architecture for image classification and improve the results on ImageNet-1K. The code will be publicly available at GitHub.

In recent years, in-vehicle applications based on vehicular ad hoc networks (VANETs) have been continuously expanded. Many applications not only focus on delay and effective forwarding rate but also pay more attention to routing path multiplexing and throughput. However, in VANETs, it is challenging to establish real-time and robust multihop forwarding paths due to volatile topological information, disconnected network, churn rate, etc. In order to adapt to the new development trend of VANETs, a self-healing routing strategy (SR) with the ant colony optimization (ACO) is proposed in this article. SR introduces the ACO algorithm to establish routing paths to ensure connectivity and immediacy. The routing-build-ability (RBA) is defined to measure the forwarding capability of a vehicle. The RBA is derived from the delay and packet delivery ratio (PDR) using the fuzzy logic system, which can reduce the computational complexity. To reduce the overhead of path reconstruction performed due to path disconnection, in-road-repairing and intersection-repairing methods are proposed in this article, which prolong the duration of the optimal path and improve throughput. The simulation results and mathematical analyses demonstrate that the feasible SR can reduce the delay by 30%, shorten the time overhead to one sixth, promote the routing duration by three times, and enhance the throughput by three times.

In this paper, we investigate the user pairing schemes for NOMA-enabled backscatter communication network. Since the successful decoding of high channel gain users can significantly improve the system performance, to make more high channel gain users be decoded successfully, we propose a pairing scheme, M-UPS, which is based on the user channel gain difference. Firstly, in the case of two-user pairing, we divide users into the high channel gain group and the low channel gain group based on channel gain, then we provide a formula of the pairing distance threshold for the high channel gain user to select low channel gain users for pairing. Secondly, we extend the proposed paring scheme to the multi-user case, a dynamic user pairing scheme, which can adaptively form the different pairs with different numbers of users based on the channel gain differences. Simulation results show that M-UPS can outperform the Conventional-NOMA (C-NOMA), Uniform Channel Gain Difference-NOMA (UCGD-NOMA), and OMA in terms of the number of paired users and total throughput.

Interference alignment (IA) is a promising technique to effectively manage the interference. The realization of IA requires a proliferation of feedback bits. In a general centralized feedback model, the feedback rate of the precoder and the decoder affects the performance of the IA directly. In this paper, to improve the feedback efficiency of the precoder and the decoder, a strategy that can adaptively allocate the feedback bits of the precoder and the decoder is proposed. We consider the effect of link loss on the throughput loss caused by the quantization error of the precoder and decoder. An upper bound of leaked interference as a function of the link loss and the feedback bits of the precoders and the decoders is derived. The feasibility conditions for dynamically allocating the feedback bits of the precoder and the decoder are analyzed. It is proven that the properties of general asymmetric interference network can satisfy the feasibility conditions for dynamic feedback scheme. According to the simulation results, our proposed scheme achieves higher throughput compared with the conventional schemes in the asymmetric interference network.

Rate adaptation, which dynamically chooses transmission rate provided at the physical layer according to the current channel conditions, is a fundamental resource management issue in IEEE 802.11 networks with the goal of maximizing the network throughput. Traditional rate adaptation algorithms for IEEE 802.11n networks do not consider the interference problem, which becomes much more serious due to the rapid deployment of IEEE 802.11n devices and large number of mobile terminals. In this paper, an interference-aware rate and channel adaptation scheme RaCA for intensive IEEE 802.11n networks was proposed. Firstly, RaCA leverages RSSI and CSI information together to measure the current channel conditions at the receiver side. RSSI is a coarse-grained indicator and CSI is a fine-grained indicator. Secondly, a two-stage rate adaptation scheme TSRA was designed, which can quickly adapt to optimal bit rate based on RSSI and CSI information. Finally, a quorum-based channel adaptation algorithm QCA was proposed, which does not need control channel. If channel suffers severe interferences, RaCA calls QCA to choose another channel to work on. Simulation and testbed implementation results demonstrate that RaCA achieves significant throughput gain over SampleLite and Minstrel-HT.

Batch-cell manipulation is a key technology in biological applications. Robotic manipulation has important significance to improve the operation success rate and reduce the technical threshold, but the problem of inefficiency still exists in batch-cell experiments. In this paper, an automated cell transportation system is designed for batch-cell manipulation. It has some technical aspects such as a cell groove to contain the cells, the micromanipulator and motor stage control methods, and computer vision algorithms. Since the cells are arranged in a line in the groove, the transportation system improves the efficiency of finding the cells in the petri dish. Furthermore, the minimum pressure to drag and release the cell are analyzed theoretically, so that the other cells will not be affected when manipulating one cell. The visual algorithms to detect the cell position and cell holding state are evaluated by porcine oocyte. Experimental results show both algorithm has high success rates: 96% and 100%. Finally, cell rotating experiments are introduced to verify the effectiveness of the transportation system. The average transfer efficiency has been improved by 20% compared to manual operation. The results show that this system can be used in many manipulations.

In this study, the authors focus on the problem of interference alignment (IA) with random vector quantisation in device‐to‐device (D2D) uplink underlaying cellular networks. For a D2D underlaying system with one cellular network and one D2D local network, they first analyse the leakage interference introduced by limited feedback and, hence, imperfect IA. Then, they derive the exact closed‐form expressions of average sum rate in terms of transmit power for cellular communication and D2D communication. Under such condition, they investigate the adaptive feedback bits allocation schemes to achieve near‐optimal performance of systems with limited feedback including a greedy feedback bits allocation scheme and a waterfilling‐based feedback bits allocation scheme. Finally, simulation results validate their theoretical results and show that significant performance gain can be obtained by allocating feedback bits adaptively.

Partial transmit sequence (PTS) is a distortionless technique used to reduce the peak-to-average power ratio (PAPR) in orthogonal frequency division multiplexing (OFDM) systems. However, it has a relatively high computational complexity. A simplified scheme is proposed in this paper. In the proposed scheme, the computation for finding the best scrambling sequences is simplified by making use of the tabu search algorithm, thus achieving the reduction of computational complexity. Algorithm analysis and simulation results show that, as compared with conventional method, the proposed scheme could be implemented with lower complexity while at no loss of PAPR performance.