Abstract:The interoperability of different assemblies is an essential factor affecting the popularization of wireless electric vehicle charging technology. Interoperability refers to the ability of the system to output rated power with specified efficiency when different ground and vehicle assemblies are paired. Aiming at the problems of incomplete evaluation criteria and low accuracy of evaluation results, this paper introduces the detuning factor and load factor representing the system resonance and load, respectively, and proposes an interoperability evaluation method based on two factors. Firstly, the limitations of the traditional evaluation method based on interface impedance are revealed. Secondly, the relationship between the detuning/load factors and system power, efficiency, and current limits are established. Then, according to the limits specified in national standards, the two factors’ interoperable region and its boundary functions are deduced, and the interoperability evaluation criterion based on the two factors is obtained. Experiments show that the proposed method can evaluate interoperability comprehensively from output power, system efficiency, and current limits. It not only solves the problem of incomplete evaluation criteria in the traditional method, but also improves the accuracy of interoperability evaluation results.

Abstract:We proposed a smartphone positioning method by formulating the positioning problem as an HMM(hidden markov model,HMM) based on the proposed double-layer feature map consisting of visual and Wi-Fi features(vision-CSI map,V-CSI map) to solve the issue of low accuracy and poor stability in indoor environment. The V-CSI map is modeled by encoding CSI fingerprint features based on grid and visual features of sparse safety exits as well as association locations. The location problem based on the V-CSI feature map is solved as HMM problem in the method. First, the safety exit sign detection and visual feature matching are completed in the visual positioning phase, and the positioning results are employed to initialize and reinitialize the states of HMM. Subsequently, CSI fingerprint features are matched with that of the V-CSI map to complete the emission probability, and the state transition probability is computed by modeling motion constraint with Gaussian model. Finally, the optimal state is derived from the forward algorithm, and the position of the smartphone is readily determined from the weighted average of the closest states. In the experiment, the proposed method is verified in an office building of 6 000 square meters and an underground parking lot of 3 600 square meters respectively. Experimental results show that the average positioning error of the algorithm is about 1.0 m, and the time of a single positioning is about 170 ms in the two typical indoor scenes. Compared with only CSI positioning methods, the average positioning error of our proposed method is reduced by more than 56%. The outstanding performance of experimental results also illustrates that our proposed method can improve the accuracy and robustness of indoor positioning.

Abstract:Traditional self-attention-based intrusion detection models have high time complexity in the calculation of attention values, and most intrusion detection models are oriented to static network environments. To address the above problems, we proposed an incremental intrusion detection model incorporating a sparse self-attention mechanism. First, we introduced a sparsity metric formula to reduce the time complexity, so as to alleviate the computational pressure of the model without affecting the detection performance of the model; Second, we constructed a dynamic example memory to alleviate the concept drift phenomenon of the model in incremental learning at the cost of a very small amount of memory space; Finally, we designed a category-balanced loss function, which is capable of enhancing the learning ability of the model for old-category samples without dynamically adjusting the model. Derivation and experiments prove that the sparse self-attention mechanism has lower time complexity and better classification effect. Compared with other schemes, the incremental learning mechanism shows a stronger ability to memorize old knowledge. The intrusion detection model has a better application prospect in the modern network environment.

Abstract:The supporting and suspension parts of the catenary are the key infrastructure of the railway catenary. However, due to the long-term contact-induced vibration between pantograph and catenary, the components of the catenary are prone to various defects. Defect monitoring based on 4C images of catenary is the key to operation and maintenance tasks, which directly relates to the safety and reliability of railway transportation. Traditional manual inspection methods face challenges such as high labor intensity, low efficiency, and a high missed detection rate. Therefore, using image processing and artificial intelligence technology to detect defects automatically is a hot issue in this research field. The components of the catenary are diverse in type, and samples of each type of defect are scarce, making the existing deep learning methods that rely on a large number of training samples difficult to apply. To overcome this problem, we proposed a classification method, named defect detection based on variational autoencoder (DefVAE) for catenary. This method was based on the assumption that samples of the same class follow a Gaussian distribution in the feature space. It utilized the potential features from the output of a variational autoencoder (VAE) to determine the feature distribution of known defect samples and generated a large amount of defect data through resampling and decoding in the distribution space to compensate for the lack of samples. During the encoding phase, we incorporated auxiliary label information to increase the inter-class distribution distance in the latent feature space. During the defect classification phase, we adopted an image generation method assisted by sliding labels and combined the reconstruction error to improve the classification accuracy. The results of comparative and ablation experiments on open-source datasets and catenary 4C datasets show that DefVAE outperforms the baseline methods in most indicators on the open-source datasets and has high classification accuracy in the classification of catenary defects.

Abstract:SAR can obtain 3D information of the target through multi-view observation. At present, multi-view SAR 3D reconstruction mainly assumes a side-looking trajectory. By constructing projective geometry equations, this type of method calculates the target offset between SAR images and derives the target height from the projective geometry. However, this type of method lacks a mathematical modeling process for the projection relationships and exhibits significant solving errors when the SAR trajectory includes squint and pitch angles. This paper analyzed the linear SAR trajectory, summarized the geometric relationship of the projection, and obtained the mathematical model of multi-view SAR projection. In the mathematical projection model, the relation matrix between pixel coordinates in the SAR imaging plane and target 3D space coordinates is called the essential matrix. The multi-view SAR mathematical model transforms the 3D reconstruction problem into a matrix inverse operation problem. The projection expression established by the essential matrix is transformed into homogeneous linear equations, and the singular value decomposition algorithm is used to solve the 3D coordinates of the target. Spaceborne SAR trajectory parameters were used for experimental simulation to verify the effectiveness of the proposed projection model and 3D reconstruction algorithm.

Abstract:In order to solve the problem of limited linearity and frame rate in the ultra-large array infrared (IR) detector readout process, this paper proposed a high-speed and high-linearity readout method. The readout circuit noise characteristics were optimized by adopting an efficient correlated double sampling (CDS) method within pixels, and the CDS voltage was output to the column bus. By employing an alternating current (AC) enhancement technique, the parasitic capacitance of the column bus was rapidly settled, while an adaptive body-bias compensation method was applied at the column bus termination to eliminate the nonlinearity introduced by the pixel source follower. A comprehensive experimental verification was conducted in the readout circuit of an 8 192 × 8 192 array IR detector based on the 55 nm process at a low temperature of 110 K. The results show that in comparison with a traditional readout circuit, the output swing is increased from 2 V to 3.3 V, and the full-well capacity is increased from 4.3 Me- to 6 Me-. The row time is reduced from 20 μs to 2 μs, and the linearity is improved from 96.9% to 99.98%. The overall power consumption of the chip is 1.6 W, and single column power consumption of the readout optimization circuit is 33 μW in the accelerated readout mode and 16.5 μW in the nonlinear correction mode.

Abstract:Heterogeneous graph neural networks have been extensively applied in data mining, information retrieval, and related domains. The metapath-based approach captures composite relationships in heterogeneous graphs by aggregating metapath neighborhood information. However, the selection of metapaths predominantly relies on prior knowledge, which may lead to the loss or overwriting of semantic information. Additionally, the use of attention mechanisms in feature aggregation incurs substantial computational overhead, and semantic confusion may arise as the network deepens or metapaths lengthen. To address these issues, a heterogeneous graph neural network that integrates multi-semantic view encoding is proposed. Firstly, all metapaths of fixed length are selected for the target node type, and subgraphs are constructed to extract corresponding semantic information. A lightweight mean aggregator is employed to obtain node representation under different metapath subgraphs, and specific relation encodings are learned for each type of metapath to combine with node representation. Subsequently, feature mapping is carried out and node features from different semantic views are fused to derive the final representation, which is applied to downstream tasks. Experiments conducted on five real-world datasets demonstrate that the proposed model more effectively captures semantic information in heterogeneous graphs, enhances node representation performance, and outperforms mainstream baseline models in node classification and link prediction tasks in most cases. The effectiveness of the model is further validated through ablation studies and parameter sensitivity analyses.

Abstract:By removing weak and redundant learners, ensemble pruning can significantly enhance the efficacy of ensemble system-based facial expression recognition. However, existing methods primarily focus on either accepting or rejecting classifiers, which results in the retention of weak classifiers or the exclusion of pivotal ones when evaluation information is unreliable or incomplete. Additionally, relying on accuracy or diversity to evaluate the merits of the classifier is difficult to fully reflect the true performance of the classifier. Consequently, this paper proposed a three-way decision-based ensemble pruning algorithm (3WDEP) for facial expression recognition, which introduced a delayed acceptance strategy to address uncertainties in classifier assessment. Simultaneously, the concept of “predictive preference” was introduced, integrating the correlation measurement between prediction results and actual labels, as well as accuracy and recall metrics, so as to construct an ensemble pruning information system and comprehensively evaluate the classifier performance. The entropy weight method was used to determine the weight of the indicators, and combined with a three-way decision, the loss of classifiers under different decision options was considered to select the classifiers that contributed the most to the ensemble system for integration. Recall was utilized as both a benefit attribute and a cost attribute to optimize the ensemble pruning effect. Experimental results show that 3WDEP effectively improves facial expression recognition performance, and the accuracy improves by 3.32%, 9.39%, 1.26%, and 4.9% compared to the initial ensemble system on FER2013, JAFFE, CK+, and KDEF, respectively.

Abstract:To address the problem of mismatch between upstream and downstream tasks exhibited by masked image modeling (MIM) methods in self-supervised representation learning, we proposed a novel pre-training model, called teacher-student complementary masked autoencoder, or in other words, the TSCAE model. The TSCAE model consists of two modules with complementary masked mechanisms, called teacher module and student module, respectively. The teacher module was designed as a Transformer-based structure to predict the masked region of an image (e.g., randomly masking 75% of the input image), while the student module employed a sole encoder to predict the remaining region of the same image (e.g., masking the remaining 25% of the input image). Meanwhile, to attain a richer visual representation from a large number of unlabeled data, the TSCAE model completed two kinds of upstream tasks, namely prediction and contrastive tasks. After that, the TSCAE model achieved the pre-training on COCO and Tiny-ImageNet datasets. The results demonstrate that across three public datasets including VOC and two private datasets, the proposed TSCAE model achieves better performance than the classical masked autoencoder (MAE) methods on downstream tasks such as image classification, object detection, and semantic segmentation. In particular, the TSCAE also alleviates the impact of the quality of the pre-training images on the visual representation learning encoder to a certain extent.

Abstract:In response to the problems of low information utilization and poor anti-noise performance in existing recognition methods for radar emitter signals of complex systems, we proposed an ensemble deep neural network recognition method integrating multiple transform domain features of radar emitter signals. Firstly, based on the three transform domain methods of bispectrum estimation, ambiguity function (AF), and Hilbert-Huang transform (HHT), we processed the emitter signals, extracted, and transformed the signal’s rectangular integral bispectrum feature, AF orthogonal slice feature, and Hilbert marginal spectrum feature into two-dimensional feature images with stronger expressiveness and interpretability. Then, we constructed a fusion recognition model framework based on ResNet18 + multilayer perceptron (MLP), took multiple ResNet18 as base learners to perform primary recognition on the datasets of three transform domain features, and obtained feature vectors represented by probabilities. Finally, we conducted fusion learning on the feature vectors via the MLP and output the final signal category information. The experimental results show that the proposed method maintains an overall average recognition rate of above 99.23% for six classes of radar emitter signals at a signal-to-noise ratio (SNR) of 0 dB. Even in the low SNR environment of -4 dB, the recognition rate remains stable at above 96.54%. The results verify the effectiveness and better performance of the proposed method.

Abstract:Dynamic cardiac magnetic resonance imaging (CMRI) is an important tool for noninvasive assessment of cardiovascular disease. In dynamic CMRI, a low-rank tensor recovery method is usually employed to explore the sparsity of dynamic magnetic resonance images; however, different modes along the tensor have different low-rank properties. The studies have found that the nonlocal self-similarity mode along the tensor can best improve the reconstruction quality of dynamic CMRI. Therefore, this paper proposes an optimal low-rank matrix recovery (OLRMR) model with matrix sparsity based on the nonlocal low-rank (NLR) method by treating each set of similar blocks extracted from a high-dimensional image as a matrix. The model uses the weighted Schatten p-norm as the rank proxy function and was solved using the alternating direction multiplier method (ADMM) and a fast soft-threshold iterative algorithm. Experimental results based on the cardiac dataset show that the OLRMR algorithm is more effective in improving the quality of the reconstructed image than the BCS, k-t SLR, and k-t LRTC algorithms and can better keep the detail of the image and the edge contour information intact. The experimental results also show that OLRMR improves the reconstruction speed by a factor of 2.6-3 over k-t LRTC.

Abstract:To significantly enhance the response speed of spacecraft formation coordination and meet strict time-window requirements for formation missions, this paper develops a novel prescribed-time performance-guaranteed control framework. This framework effectively addresses three critical challenges prevalent in formation systems: limited perception range, actuator saturation constraints, and inter-agent collision avoidance requirements. First, by integrating error transformation technique with sliding mode control, a control Lyapunov function condition is constructed. This design not only ensures the system meets strict timing requirements for formation tasks, but also guarantees prescribed transient response characteristics and steady-state performance metrics. Second, through the establishment of high-order control barrier functions, precise regulation of relative distances between adjacent spacecraft is achieved, maintaining formation communication topology connectivity while effectively preventing collision risks. Furthermore, this study employs quadratic programming to solve for optimal control inputs, realizing multi-objective coordinated optimization of prescribed-time convergence, topology connectivity maintenance, and collision avoidance control under actuator saturation constraints. To validate the effectiveness of the proposed control framework, systematic performance verification is conducted through numerical simulations. The simulation results fully demonstrate the reliability and superiority of the proposed control scheme in satisfying all specified constraints.

Abstract:Video re-localization aims to localize a moment that semantically corresponds to a given query video from an untrimmed reference video. This task not only meets the actual browsing needs of users but also plays an important role in various application scenarios. Since videos contain richer information compared to other data forms like images and text, accurately identifying the target moment in a long video and determining its temporal boundaries are significantly challenging. This paper regarded the video re-localization task as a sequential decision-making process and applied reinforcement learning to achieve efficient and accurate localization. Specifically, this paper proposed an agent-guided localization network (AGLN), which trained an agent to progressively refine temporal boundaries of the localized moment based on the learned policy, thereby finding the most relevant moment to the query video. Additionally, AGLN combined reinforcement learning with supervised learning in a multi-task learning framework, aiding the agent in more effectively exploring the environment and learning the optimal policy. Experimental results on the ActivityNet-VRL dataset demonstrate that AGLN outperforms existing methods in the video re-localization task. The average retrieval accuracy of AGLN is 25.9%, which is 0.2 percentage points higher than the current optimal method.

Abstract:The existing all-symbol locally repairable codes with unequal availability have limited parameter values, and low code rates. In order to solve the problems above, this paper constructs a class of all-symbol locally repairable codes with unequal availability based on saturated orthogonal arrays, which achieves higher code rates. Specifically, the association matrix is generated based on the saturated orthogonal array, and the matrix transformations as well as the Kronecker product operations are performed on the association matrix to generate all-symbol locally repairable codes with high availability of information symbols and unequal availability of information bits. Theoretical analyses show that all-symbol locally repairable codes with high availability of information symbols have flexible parameter values and are optimal in dimension and code length at locality r=2. Compared with the existing all-symbol locally repairable codes, the constructed all-symbol locally repairable codes with unequal availability of information bits perform better in terms of code rate.

Abstract:To enhance the microstructure and mechanical properties of laser cladded coatings, this study focuses on Ni60/WC coatings and proposes a resistance heating heat treatment (RHHT) process using pulsed direct current. First, RHHT experiments were carried out on Ni60/WC coatings for 1 h and 2 h, respectively, under the current density of 3.33 A/mm² .Subsequently, SEM and XRD were used to analyze the phase composition and microstructure transformation of coatings, and mechanical properties of the specimens before and after RHHT were tested. The results show that due to the selective heating effect, the electric current bypasses the hard phases within the coating, generating localized high temperatures which cause the secondary decomposition of WC. The Ti, Cr, and C atoms dissolved in γ-(Ni, Ti) diffuse under the influence of electric current, and the phase transformation in the coating proceeds in the direction of increasing electrical conductivity. Moreover,due to the electric current reducing the nucleation energy barrier in combination with rapid cooling, the nucleation rate was significantly increased and the grain sizes were reduced by approximately 99% which resulted in fine γ-(Ni, Ti) grains after the RHHT. By comparing the mechanical properties of the coatings before and after RHHT, it was found that the microhardness, fracture toughness and wear properties were effectively improved. The direct current, through its selective heating effect and athermal effects, circumvented the grain defects, and enhanced atomic diffusion ability, increased the nucleation rate, and refined the grains in the coating.

Abstract:Existing rumor detection methods mainly rely on text semantic features and network propagation features, but the source tweets dominated by short texts can easily lead to insufficient semantic features, and the propagation tree used to extract propagation features can generate a large amount of data. To solve these problems, we proposed a rumor detection method, namely multi-view graph neural network, which integrated rich semantics and global propagation. This model used source texts to get structural semantic relationships, utilized external knowledge to extract potential semantic relationships, and got the global propagation relationship among users by source tweets and their response users. Finally, it automatically learned the feature weights of different views through the attention fusion mechanism, achieving adaptive information fusion and improving the accuracy of rumor detection. Besides, it adopted Word2Vec to supplement the content semantics of source tweets. Experimental results show that using source texts, external knowledge, and response users to construct graphs, respectively, can effectively capture rich semantic information and concise global propagation relationships. The model outperforms a series of baseline models on the public datasets Twitter15 and Twitter16, with the accuracy rates of 90.2% and 90.8%, respectively. The analysis results from the ablation experiment show that the proposed method can comprehensively capture rich semantic features of the source tweets and effectively obtain the global propagation relationship in a concise manner, so as to improve the accuracy of rumor detection.

Abstract:Hyperspectral image (HSI) classification is a challenging task in the field of remote sensing, because the HSI has high spectral dimensionality and low spatial resolution, which makes it difficult to fully extract the spatial-spectral features of hyperspectral images in the classification task. Aiming at solving the problems of the existing convolutional neural network (CNN)-based HSI classification models, such as large parameter size, high computational cost and low classification accuracy, a lightweight network hyperspectral image classification model based on attention mechanism (AMLW-CNN) is proposed in this paper. In order to enhance feature extraction ability of the network, the spatial-spectral feature extraction module is designed based on two multiscale extraction modules. In addition, we use the residual structures to connect the convolutional layers of spatial feature extraction module and incorporate the attention mechanism to enhance the extraction of useful features. Furthermore, to reduce the number of model parameters, an asymmetric convolution and a depthwise separable convolution are introduced to replace the 3D and 2D convolution kernels, respectively. The experimental results show that classification accuracy of AMLW-CNN is better than that of the comparison algorithms, with lower computational complexity and higher robustness. The overall classification accuracies on the datasets of Indian Pines, Salinas and Pavia U has attain 98.5%,99.8% and 99.9%, respectively.

Abstract:To reduce the risk of instability and wrinkling in the reducing and thickening extrusion process for aluminum alloy pull rod, and improve the production efficiency of the pull rod, a non-isothermal hot extrusion forming process was proposed, in which the fixture clamps the non-deformation zone of the tube blank, with only the extrusion die preheated and no cooling required for the tube blank. Firstly, the non-isothermal hot extrusion forming process of 2A12 aluminum alloy pull rod with diameter of  65 mm, thickness of 5 mm and target thickening ratio of 1.6 was analyzed based on finite element simulation, and then the effects of extrusion process parameters on forming quality were investigated. Secondly, under the optimized process parameters determined by simulation, the non-isothermal hot extrusion experiment of the pull rod was conducted, validating the effectiveness of the numerical simulation. The study demonstrated excellent agreement between simulation results and experimental data. The non-isothermal hot extrusion process eliminated bending deformation at the tube ends. Furthermore, the gradient-decreasing temperature distribution at the leading end of the tube along the extrusion direction helped maintain the material strength in the non-deformation zone, thereby reducing the risk of instability and wrinkling. Under the extrusion process parameters of friction coefficient of 0.05~0.3, extrusion speed of 1.6~7.6 mm/s and die temperature of 410~470 ℃, the friction coefficient exhibited a significant influence on the forming quality of the pull rod, while the extrusion speed and die temperature had relatively minor effects. However, none of these parameters showed a notable impact on the wall thickness uniformity in the thickened zone of the pull rod. As the friction coefficient decreased, the risk of wrinkling of the pull rod decreased, and the wall thickness of the thickened zone increased with a maximum increase of 56.5%. The findings of the study offer an innovative methodology for enhancing the extrusion forming quality of aluminum alloy pull rod.

Abstract:The permafrost in northeastern China constitutes an important component of the Xing’an-(trans)Baikal permafrost, exhibiting characteristics of both high latitude and high altitude permafrost. The presence and dynamics of permafrost directly impact the ecological environment of the regional cold zone, water-carbon cycles, cold region engineering design, and operations. Currently, empirical and semi-empirical models based on thermal boundary conditions tend to overestimate the areal extent of permafrost and insufficiently consider environmental factors beyond temperature. To more accurately delineate the regional distribution of permafrost, this paper obtained the spatial variation characteristics of zonal and non-zonal factors influencing regional permafrost distribution through regional surveys and data integration and employed the boosted regression tree model for simulation and analysis. The results indicate that among the zonal factors, latitude, longitude, and altitude contribute 45.3%, 42.4%, and 12.3%, respectively. Among the non-zonal factors, temperature (including freezing and thawing indices), precipitation, soil-water conditions, snow cover, and vegetation contribute 46.4%, 18.9%, 13.1%, 12.5%, and 9.1%, respectively. This analysis clarifies the contributions of environmental factors to the development and dynamics of Xing’an-(trans)Baikal permafrost. Compared with a classification and regression decision tree, the boosted regression tree model achieves an accuracy of 0.91. This study provides data support and reference for regional permafrost research and related fields.

Abstract:In the complex application environment of wireless sensor networks (WSN), in order to resist the selective forwarding attack and dishonest recommendation attack launched by malicious nodes and improve the safety performance of the network, this paper proposes a malicious node identification strategy based on artificial bee colony (ABC) considering reputation (CR-ABC) in WSN. By utilizing a fuzzy trust model (FTM) and integrating a fuzzy comprehensive evaluation mechanism, the paper calculates the comprehensive trust value of nodes based on three influencing factors: communication features, data attributes, and physical attributes to improve the detection accuracy of the reputation model. The paper introduces the suggested deviation function and the interaction index deviation function and uses the ABC algorithm to optimize the FTM, aiming to ensure that the system still maintains a higher identification rate and a lower misjudgment rate when there are too many malicious nodes. The simulation results show that the identification rate of CR-ABC for selective forwarding attacks can reach over 90%, and the misjudgment rate for normal nodes can be reduced to less than 6%. For dishonest recommendation attacks, even if the number of dishonest nodes reaches 50%, CR-ABC still maintains a high identification performance, which can effectively improve the security and reliability of WSN in complex environments.

Abstract:The face image inpainting technology can generate a complete face image by repairing the occluded area of the face, which has important application value in fields such as criminal investigation and security protection. However, the inpainting results of the existing methods often exhibit artifacts such as fuzzy texture and distorted face structure. Therefore, based on the generative adversarial network (GAN) framework, this paper proposed an occluded face restoration network fusing edges and key points. Firstly, the proposed network used the structural forest edge restoration network to complete the structural forest edge map occluding the face image to obtain more description information of the face details. Then, it used the key point prediction network to locate 68 key points of the occluded face to obtain the topological structure information of the face image. Finally, it took the structural forest edge map and the key points of face obtained by the above two networks as prior information, restored the occluded face area by the face image inpainting network, and generated a complete face image. The experimental results on the CelebA-HQ dataset show that the face images restored by the proposed algorithm have finer texture details and more reasonable topological structures of faces. Under different occluded areas, the PSNR and SSIM of the proposed algorithm are higher than those of the comparison algorithm. Compared with that of GatedConv, EdgeConnect, and LaFIn algorithms, when the mask ratio is 50%, the PSNR of the proposed algorithm increases by 36.8%, 25.8%, and 29.3%, respectively, while the SSIM increases by 19.5%, 12.2%, and 12.2%.

Abstract:With the advent of the era of big data, network structures are becoming more and more complex, and exploring the community structure of complex networks holds great significance in understanding their function and organization mechanisms. Many studies have been conducted for community detection, among which the label propagation algorithm (LPA) has a near linear time complexity and is applicable for large-scale complex networks. However, it has excessive randomness and relatively low accuracy. This paper proposed a collective influence-based label propagation algorithm (CILPA) for discovering overlapping community structures. CILPA introduced collective influence as a global indicator, redefined the node importance by integrating the node’s own information and global network information, and fixed node update order according to node importance to improve the algorithm’s stability. In the label propagation process, a label selection strategy was designed, and the adaptive filtering factor were set to prevent the interference of wrong labels, thereby improving the accuracy and robustness of the algorithm. Finally, experiments were conducted on artificial and real networks with different scales, complexities, and overlap rates. The results show that the modularity and normalized mutual information of CILPA are superior to those of mainstream algorithms such as COPRA and SLPA, with a smaller standard deviation. This indicates that the proposed method possesses both effectiveness and stability in overlapping community detection, providing a reliable method for the analysis of overlapping communities in large-scale complex networks.

Abstract:In order to address the challenge of designing reasonable offloading decisions for mobile edge computing (MEC) in multi-user environments, which leads to load imbalance, excessive total latency, and response delays, this paper proposed a latency-sensitive heuristic task offloading method. Firstly, to address the issues of limited computational resources and insufficient battery power of edge devices during computation task processing, the paper introduced an edge server-centric offloading paradigm and established a system model and a latency optimization model. Subsequently, it introduced an improved proximal policy optimization algorithm (I-PPO), which extended the offline training process, designed a reward mechanism that considers the impact of multi-agent decisions, and incorporated global information based on specific agents into the features, enabling the algorithm to be suitable for multi-user environments. Furthermore, building upon I-PPO, the paper introduced task priority scheduling decisions into the task offloading execution process, resulting in the development of a latency-sensitive lightweight heuristic task offloading algorithm, denoted as HTAI. This further optimized system latency and enhanced user satisfaction. Simulation experiments demonstrate that the I-PPO algorithm proposed in this paper, compared to similar algorithms, effectively improves convergence speed, optimization capability, and robustness, and it can be applied in multi-agent environments. Moreover, the algorithm proposed herein outperforms other algorithms in terms of total system latency and edge server load balance, exhibiting strong stability.

Abstract:In order to realize the demand for low-cost and long-term detection of large-scale building groups and to expand the identification range of the sensor, this paper proposed a metal crack sensor based on passive chipless radio frequency identification (RFID) technology. Based on the influencing factors such as cross-polarization and operating bandwidth and a large amount of simulation data on the HFSS platform in the early stage, the paper designed a sensor model with excellent detection performance. Horizontal, vertical, and diagonal cracks were constructed, and electromagnetic excitation of the plane wave was used to test the influence of different crack shapes on the sensing and detection. The position of various types of cracks was changed, and changes in the electric field of the resonant cavity, current, and response amplitude were analyzed to determine the optimal identification range of the sensor. The results show that the average amplitude deviation of the sensor’s response to structural damage detection is 5 dB in the ultra-high frequency band. The change of crack position will affect the surface current distribution, which will change the response amplitude, while the structural damage response is detuned compared with the crack-free response, and the change of crack position does not affect the detectability of cracks. The sensor is capable of detecting cracks in different directions at any position on the surface of an object over a full range, improving the identification range and enabling real-time monitoring of cracks with small positional variations at a high resolution.