Search

In response to the harsh working conditions of high-power external rotor permanent magnet synchronous motors, a water-cooling structure is utilized to balance the heat generated by an external rotor permanent magnet synchronous motor, ensuring effective heat dissipation. In this study a two-dimensional electromagnetic simulation model was established to obtain a cloud diagram illustrating the temperature field distribution of a permanent magnet motor using the magnetic-thermal coupling method. Three cooling water channel models were designed: the axial Z-type, rectangular water channel opening circumferential spiral type, and circular water channel opening circumferential spiral type. Additionally, the temperature field distribution of the motor windings during stable operation is obtained using the electromagnetic field-fluid field-temperature field coupling method. By comparing different cooling schemes, the optimal cooling system was selected, which showed a 32.57% decrease in the winding temperature after the implementation of the water-cooling system. The water-cooling system designed in this study has some theoretical and practical value in the study of heat dissipation systems for external rotor permanent magnet synchronous motors.

The distribution network line loss computation method needs to be enhanced in light of the ongoing growth of the national power grid. The study classifies and segments the station data using a decision tree model and a multi-feature volume weighted station clustering algorithm. It then uses a back propagation neural network as a substrate, along with the Levenberg-Marquard algorithm and genetic algorithm for optimization. Collect and organize relevant data on line loss rates in low-voltage substation areas, including information on energy meters, meter boxes, and lines. Next, construct a genetic algorithm neural network model and use the backpropagation algorithm for training. Evaluate the accuracy and stability of various models by comparing the error between predicted and actual line loss rates through experiments. Finally, optimize the neural network parameters and network structure to improve the model’s prediction accuracy and robustness. The experimental data showed that compared to the density-based spatial clustering algorithm for noisy applications, the contour coefficient metrics of the proposed multi-feature volume weighted station clustering algorithm improved by 0.05 and the average consumption time of the algorithm was reduced by 75%. Compared to the back-propagation neural network model optimized by the Levenberg-Marquard algorithm, the root-mean-square error of the neural network model optimized by the addition of the genetic algorithm for the calculation of the line loss rate of the four station samples was reduced by 72%, 55%, 53% and 37%, and the values of R2 were improved by 8.72%, 13.59%, 7.91% and 11.69%, respectively. The testing results demonstrated that the neural network model has good generalization capabilities and a high degree of curve fitting. Also, the relative errors of the calculation of the station area’s line loss rate are mainly within the range of 0% and 10%. For the growth of energy conservation in the country, this innovative technology offers a new way to determine and manage line loss of the station area.

Deep learning theory has made remarkable advancements in the field of baby cry recognition, significantly enhancing its accuracy. Nonetheless, existing research faces two challenges. Firstly, the limited size of the database increases the risk of overfitting for a deep learning model. Secondly, the integration of multi-domain features has been neglected. To address these issues, a novel approach called BCRNet is proposed, which combines transfer learning and feature fusion. The BCRNet model takes multi-domain features as input and extracts deep features using a transfer learning model. Subsequently, a multilayer autoencoder is utilized for feature reduction, and a Support Vector Machine (SVM) is employed to select the transfer learning model with the highest classification accuracy. Then two features are concatenated to form fused features. Finally, the fused features are fed into a deep neural network for classification. Experimental results show that the proposed model is effective in mitigating the model overfitting problem due to small datasets. The fused features of the proposed method are better than the existing methods using single domain features.

Assessing fall risk accurately is vital for the older adult population. However, existing fall risk assessments mainly depend on scales, which are inconvenient, subjective, and imprecise. The aim of this study was to explore a machine learning model based on gait parameters to evaluate the risk of falls in older adults living in the community over a one-year period. A total of 46 elderly subjects were recruited in this study. Information on demographics, disease history, and fall history was collected via questionnaire. Moreover, this study used a gait analysis system based on inertial measurement unit and Azure Kinect to acquire the spatiotemporal parameters of the subjects’ gait. Based on the above data, various machine learning models, including k-nearest neighbor, support vector machine, gradient boosting decision tree, and voting classifier, were built to estimate the fall risk level of elderly individuals. K-nearest neighbor performed best among all the models with an accuracy of 0.80 on the individual test set, an F1 score of 0.67, and an area under the receiver operating characteristic curve of 0.83. Gait frequency was found to be the most significant feature associated with fall risk, followed by body mass index and gait cycle variability. The findings suggest that the k-nearest neighbor model can provide a quantitative and objective evaluation of fall risk for older adults living in the community and that the evaluation is more accurate when both gait parameters and disease history are taken into account.

In the global outbreak of corona-virus disease (COVID-19), it is of foremost priority to find an efficient and faster diagnosis method to reduce the transmission rate of the COVID-19 disease. Recently, deep learning-based detection methods have been proposed as a potential diagnostic tool, because there is much essential information in the radio-logical images for COVID-19 detection. This paper intends to develop a hybrid deep neural network (COVID19-ResCapsNet), using the X-ray images, to predict the risk at the onset of disease in patients suffering from COVID-19. The proposed COVID19-ResCapsNet method consists of two steps. The first step is mainly to extract features with the improved Residual feature extraction network. In the final step, several Capsule Networks are employed to classify COVID-19, and Non COVID-19. The proposed method is evaluated on two CXR image datasets(Dataset-1: 4236 images and Dataset-2: 2250 images) and the classification accuracy can reach 0.9988 on Dataset-1 and 0.9933 on Dataset-2 for binary classification cases. The results are compared with several state-of-the-art pre-trained classification models. The effects of two important hyper parameters: different optimizer and batch size, have also been investigated. Hence, the proposed COVID-19 detection model is effective and can be used as the supplementary tool in clinical application.

In this paper, we propose parallel heuristic methods to accelerate the generation of $(n,m)$ best equivocation code (BEC), where $n$ and $m$ are code and message lengths, respectively. The proposed dynamic programming (DP) method and greedy method extend a previous heuristics method by reducing the time complexity of the code generation process. The DP method produces the same codes as the previous method but incurs an overhead for data reuse. In contrast, the greedy method avoids this overhead but generates slightly different codes due to its heuristic approach. We parallelize the proposed methods by exploiting coarse-grained and fine-grained parallelisms, which achieve further acceleration on multicore CPU and graphics processing unit (GPU) systems, respectively. Experimental results demonstrate that the proposed DP and greedy methods reduce the sequential generation time to a quarter, as indicated by theoretical complexity analysis. In addition, the parallel implementation achieves linear speedup on a multicore CPU system, and the GPU implementation realizes coalesced memory accesses, resulting in $17\times $ acceleration over the eight-core CPU implementation. We found that the greedy method produced different codes that differ from the previous and DP methods; however, the generated codes had higher equivocation rates than those generated by a naive random method. We believe that the proposed parallel methods can effectively accelerate BEC generation for large $m$ and $n$ values, especially with larger values of $n$ relative to $m$ .

The development of electrification has put forward higher requirements for the efficiency and power density of permanent magnet motors. The combination of formed flat wire windings and open slots with rectangular cross-sections has been applied in permanent magnet motors, which makes the motor torque fluctuate significantly, and the vibration and noise caused by it seriously influence the stability of operation. Therefore, it is essential to study the torque characteristics for permanent magnet motors with formed windings. The open slot brings rich tooth harmonics to the air-gap magnetomotive force. In this study, an analytical model of the motor magnetic field was established based on magnetic potential and relative permeance theory. According to the Lorentz force law, an analytical model of the instantaneous torque of the permanent magnet motor with a formed winding is derived. The model points out that under the modulation of the relative permeance function, the constant torque and torque fluctuation are generated after the interaction of the stator magnetomotive force and rotor magnetic potential that meet the corresponding harmonic order relationship, and the order of the torque fluctuation is a multiple of 6. According to the analytical model, different torque components of the permanent magnet motor with formed windings are calculated and verified by the finite element method.

In this paper, a compact, band-notched ultra-wideband (UWB) fully-recessed antenna with pattern diversity is proposed and demonstrated for vehicle-to-X communications. The antenna is composed of a multi-port central radiator fully embedded inside a bowl-shaped ground structure with a flush top. The multi-port central radiator consists of a top circular plate, four split-quarter conical monopoles, four vertically- and horizontally-orientated strips, a central metallic cylinder, and a bottom cylinder. Moreover, four F-shaped monopoles are loaded to the multi-port central radiator for producing two narrow notched bands. The integrated antenna achieves a form factor of $0.23 \lambda_L^3$ , where $\lambda_L$ is the free space wavelength at the lowest working frequency. The proposed antenna was fabricated and characterized, achieving a measured $\mathrm{S}_{11}\lt -10 \mathrm{~dB}$ band ranging from 1.7 to 6 GHz with two stopbands at around 2.5 and 5.15 GHz for all four ports. Moreover, the measured mutual coupling is smaller than −15 dB, while the envelope correlation coefficients (ECCs) are below 0.1 between different input ports. A vertically-polarized (VP) omni-directional pattern can be supported when all four ports are excited simultaneously, whereas four directional patterns pointing towards different directions in the azimuthal plane can be realized when the four ports are excited individually. The demonstrated antenna is a promising candidate in multiple-input multiple-output communication systems for a wide variety of platforms.

This article presents a comprehensive evaluation of the 2000–2018 Clouds and Earth's Radiant Energy System Synoptic 1° Ed4A (CERES SYN1deg Edition 4A) surface solar radiation (SSR) product. In particular, the global assessment is conducted over different temporal scales (i.e., hourly, daily, and monthly-average) with special attention given to the impact of clouds, and a regional evaluation is further implemented over the Mainland of China (MC) using SSR measurements from a denser observational network provided by the China Meteorological Administration. Evaluation across all valid station-grid pairs yields mixed performance with |MBE|≤2.8 (6.2) W m−2, RMSE≤89.5 (31.6) W m−2, and R≥0.95 (0.93) over the globe (MC) for different temporal scales, and the monthly CERES SSR, with RMSE≤20 W m−2, is found to hold promise for global numerical weather prediction and climate monitoring. In addition, CERES is found to generally underestimate and overestimate SSR over land and ocean, respectively. Comparison between year-round and cloudy-season suggests that the presence of clouds may potentially impact the SSR retrievals, especially at the hourly temporal scales, with an increase in RMSE values larger than 10 W m−2 for most stations. Further investigation of subgrid heterogeneity suggests that most in situ SSR measurements can reasonably represent the 1° grid average except for some stations with specific geographic deployments, which may raise significant spatial representativeness issues and, therefore, need to be used with great caution.

This article investigates a novel cooperative control system based on the sliding mode variable structure control theory for multi-missile formation flight. It is desired in practice to form and maintain the formation under the premise of uncontrollable missile speed. First, under the inertial coordinate system, we obtain the model of the formation control problem using the relative position of the leader to the follower. Then, we perform acceleration conversion and combine it to this model, getting the form of the formation problem model in the ballistic coordinate system. The finished model is useful for researchers to design the formation controller on this basis. Besides, the sliding mode variable structure control theory is used to design the formation controller for the system not considering disturbances and considering disturbances. Then we use the Lyapunov stability theory to analyze the stability of the formation control system. Finally, we compare our method with another method which requires controllable speed. According to numerical simulations, the method proposed in this article can achieve similar relative position errors under the condition of uncontrollable speed. And the robustness, versatility and formation adaptability of the method we propose are confirmed by simulation results.

Multi-view based 3D reconstruction aims to obtain 3D structure information of objects in space through two-dimensional images. In this paper, we propose a new multi-view stereo network that can robustly reconstruct the scene. To enhance the feature representation ability of Point-MVSNet, a pyramid attention module is introduced. Specifically, we exploit the attention mechanism for the multi-scale feature pyramid to capture larger receptive fields and richer information. Instead of constructing a feature pyramid as the input, results of the pyramid attention module at different scales are directly used for the next layer. The network eventually generates a high-quality depth estimation for 3D reconstruction from sparse to dense by an iterative refinement schedule. Experiments have been performed to evaluate 3D reconstruction quality by comparison with existing state-of-the-art methods on the DTU dataset. The experimental results indicate our method performs the best in overall quality compared with previous methods, proving the effectiveness of our method. In the end, we use the data collected by mobile devices to implement 3D reconstruction with a combination of traditional and learning-based methods, providing ideas for the 3D reconstruction technology on mobile devices.

In social media, images and texts are used to convey individuals' attitudes and feelings; thus, social media has become an indispensable part of people's lives. To understand social behavior and provide better recommendations, sentiment analysis on social media is helpful. One sentiment analysis task is polarity prediction. Although current research on visual or textual sentiment analysis has achieved quite good progress, multimodal and cross-modal analysis combining visual and textual correlation is still in the exploration stage. To capture a semantic connection between images and captions, this paper proposes a cross-modal approach that considers both images and captions in classifying image sentiment polarity. This method transfers the correlation between textual content to images. First, the image and its corresponding caption are sent into an inner-class mapping model, where they are transformed into vectors in Hilbert space to obtain their labels by calculating the inner-class maximum mean discrepancy (MMD). Then, a class-aware sentence representation (CASR) model assigns the distributed representation to the labels with a class-aware attention-based gated recurrent unit (GRU). Finally, an inner-class dependency LSTM (IDLSTM) classifies the sentiment polarity. Experiments carried out on the Getty Images dataset and Twitter 1269 dataset demonstrate the effectiveness of our approach. Moreover, extensive experimental results show that our model outperforms baseline solutions.

The grounding grid is the cornerstone for maintaining the safe operation of the power system. The grounding network acts as a concealed project, and its invisible characteristics make the defects difficult to be detected, which causes major safety hazards in the system. Thus accurate diagnosis of grounding grid corrosion is vital in electrical engineering. A diagnosis method for grounding grid corrosion based on branch voltage disturbance is proposed in this paper, which can judge the location and degree of corrosion through branch voltage disturbance before and after corrosion. Firstly, the sensitivity of branch voltage to branch resistance of the grounding grid is analyzed. Secondly, a diagnostic model based on branch voltage disturbance is established, and the corresponding regularity between peak voltage disturbance and the corrosion state of the grounding grid branch is developed. Finally, it is verified by experiments that this method can effectively diagnose the corrosion of the grounding grid.

Numerous IoT applications have emerged in human healthcare area with advances in wearable electronics. Multiple physical and physiological data bearing strong spatial-temporal characteristic collected by the wearable sensors is sent to the smartphones where it is aggregated and transferred to back-end applications for further processing. Analyzing the multivariate time series is of great importance yet very challenging as it is affected by many complex factors, i.e., dynamic spatial-temporal correlations and external factors. In this paper, we propose an attention-based encoder-decoder framework for multi-sensory time-series analytic. It consists of four parts: data collection, data mining, time-series analytic part and user interaction. A temporal-attention based encoder-decoder model is proposed to make a long-term prediction of multiple time series to realize the real-time user interaction. The proposed model uses the LSTM model to learn the long-term dependence of the time series related to certain motion sequence. The attention mechanism connects the encoder and the decoder to make long-term predictions for future time series. Through extensive experiments, the proposed model has achieved better results in short-term and long-term predictions compared with the state of art methods. An activity recognition algorithm based on LSTM is also proposed in this framework to identify daily human activities and sports activities accurately. Through five-fold and ten-fold cross-validation strategies and comparison with six baseline machine learning models, the activity recognition algorithm has a recognition rate of 98.89% and 99.28% for human activity.

The time interval of the observational data changes irregularly because of the difference of sensors’ sampling rate, the communication delay and the target leaving observation region of the sensor sometimes. These problems of asynchronous observation data greatly reduce the tracking accuracy of the multi-sensors system. Therefore, asynchronous data fusion system is more practical than synchronous data fusion system, and worthier of study. By establishing an asynchronous track fusion model with irregular time interval of observation data and combining with the Track Quality with Multiple Model (TQMM), an asynchronous track fusion algorithm with information feedback is proposed, and the TQMM is used for weight allocation to improve the performance of the asynchronous multi-sensor fusion system. The simulation result shows that the algorithm has better tracking performance compared with other algorithms, so that this kind of problem of track-to-track fusion for asynchronous sensors is solved effectively.

Under the high load, high frequency and high strength operating environment, the frequent occurrence of vehicle fault gradually attracts the attention of the society. The real-time monitoring and data recording function of vehicle-mounted equipment provides data support for vehicle status assessment and fault warning. In this paper, the real-time data collected by CAN-BUS system of Beijing Bus Group are preprocessed and discretized. On the basis of the traditional rough set theory, a new coding method is set up, and the dependency between conditional attributes and decision attributes is set as an adaptive function, which is reduced by genetic algorithm and cellular genetic algorithm respectively. The calculation results show that the key fault information of public transport vehicles is instrumental speed, oil pressure, percentage of torque, timing engine speed, and coolant temperature. By comparing the results of reduction, it is found that the cellular genetic algorithm has higher applicability than the genetic algorithm in terms of algorithm efficiency, stability, and convergence quality. Although the genetic algorithm attribute reduction is slightly better than the cellular genetic algorithm attribute reduction in the rule matching, the cellular genetic algorithm has a better ability to excavate information within the acceptable compatibility range. Finally, the selected key factors will be deployed on the Beijing Bus Group's big data platform and displayed in real time. The conclusion of this paper enriches the theory of bus engine fault warning and establishes an engine failure warning system, which can effectively reduce the failure rate of bus vehicles and reduce the maintenance cost expenditure. It has certain guiding significance for the bus operation work of Beijing Bus Group.

The subway station passenger flow prediction model can forecast passenger volume in the future. This model helps to carry out safety warnings and evacuation of passenger flow in advance. Based on the data of the Shanghai traffic card, the passenger volume in all the time intervals is clustered into three different models for prediction. Taking the Nanjing East Road Station in Shanghai as an example, the time series of passenger volumes was combined with weather data to create several supervised sequences and was converted to supervised sequences according to different values of timestep. To accelerate convergence, two artificial features were added as input. The gated recurrent unit (GRU) network model achieves accurate rolling prediction from 15 minutes to 6 hours. Finally, comparing it with the long short-term memory (LSTM) network and the back-forward propagation network (BPN), it was confirmed that the GRU network with a timestep of 1.5 hours is the best model for the long-term (more than 3 hours) traffic flow rolling prediction, while GRU with a timestep of 45 minutes has the best result for short-term rolling prediction.

Schizophrenia is a devastating disease with a prevalence of 1% in populations around the world. Current diagnostic techniques of schizophrenia and high-risk population are based on subjective psychiatric interviews. Early diagnosis and intervention can mitigate progression and improve treatment outcomes. However, the lack of biomarkers that support objective examinations has been a long-term bottleneck in clinical diagnosis and assessment of schizophrenia and its high-risk state. In the present study, resting-state 128-channel electroencephalogram (EEG) data were acquired from 65 participants, including clinically-stable individuals with first-episode schizophrenia (FESZ), individuals at ultra-high-risk (UHR) and high-risk (HR), and healthy controls (HC). Microstate analysis was used to assess the dynamics of functional networks in these participants. Three features were extracted for each class of microstate (A, B, C, D, E, F): duration, occurrence and time coverage. Furthermore, clinical examinations and cognitive tests were performed. Behavioral results showed poorer performances in the participants as the disease progressed. Moreover, microstate features computed from resting-state EEG microstates (especially microstate class D) were capable of distinguishing the four groups of individuals. Combined biomarkers including clinical examinations, cognitive tests and EEG microstate parameters were identified as a potential effective diagnostic tool, achieving the highest classification performance using the random forest model compared with the support vector machine (SVM) and long short term memory (LSTM) networks, with an average classification of 92%, mean sensitivity of 91.8%, and specificity of 90.8% among the four groups, which were much higher than that only using behavioral features. The results demonstrate that microstate-based indicators together with behavioral results may act as biomarkers for early diagnosis and prediction of at-risk individuals of schizophrenia. Furthermore, our findings illustrate the potential use of resting-state EEG in clinical screening, classification and quantitative evaluation of patients with neurodevelopmental disorders.

Imbalance learning is a hot topic in the data mining and machine learning domains. Data-level, algorithm-level and ensemble solutions are the three main methods proposed thus far to address imbalance learning. To alleviate the issues of data explosion and feature selection in multilayer perceptron based on simultaneous two-sample representation(S2SMLP), in this paper, firstly, spectral clustering is exploited to select majority samples so as to construct a smaller training dataset for the classifier. We divide all majority samples into many clusters through spectral clustering, extract different numbers of representative samples from a cluster according to the size of each cluster, the average distance between the minority class and all samples of the cluster, then construct the training dataset of the classifier by combining these extracted samples from the majority class and all minority samples. Secondly, we propose a novel feature selection method based on the pairwise samples distance constraint, which considers the class labels of paired samples, select the features which push two similar samples closer together and pull two different samples farther apart. Finally, we conduct extensive experiments on 44 two-class imbalanced datasets and four high-dimensional DNA microarray datasets. The experimental results demonstrate that our proposed algorithms outperform some state-of-the-art algorithms in terms of $F\textrm {-measure},G\textrm {-mean}$ and AUC.

In recent years, correlation filter (CF)-based tracking methods have demonstrated competitive performance. However, conventional CF-based methods suffer from unwanted boundary effects because of the periodic assumption of the training and detection samples. A spatially regularized discriminative CF (SRDCF) has greatly alleviated boundary effects by proposing the spatial regularization weights, which penalize the CF coefficients during learning. However, the SRDCF utilizes a naive decaying exponential model to passively and fixedly update the CF from the previous results. Therefore, if the target meets with occlusion or is out of view, the SRDCF may encounter over-fitting to the recent polluted samples, which may lead to tracking drift and failure. In this paper, we present a novel CF-based tracking method to resolve this issue by dynamically and adaptively correcting the weights of learning CFs and fusing them together to promote a more robust tracking. Thus, if the recent samples are inaccurate in the case of occlusion or are out of view, our method will down-weight the corresponding CFs and vice versa. Moreover, in order to decrease computational complexity and ensure memory efficiency, we extract the key CFs from the previous frames to remove redundant CFs under the contiguous frame indexes constraint. Thus, we do not need to store all CFs and decrease computational burden. Benefiting from the extraction and enhancement of CF, our method improves the tracking precision on OTB-2015, VOT-2016 and UAV123 benchmarks and achieves a 56.0% relative gain in speed compared with the SRDCF. The extensive experimental results demonstrate that our method is competitive with the state-of-the-art algorithms.

The content-based video retrieval is a popular topic in computer vision filed, especially, action retrieval. This paper proposes a novel and effective video representation module for content-based action retrieval framework, in which feature learning can be conducted with complementary information and intrinsic structure, where the relationship between appearance feature and geometry can be preserved. Based on multi-view analysis and graph embedding, the target features are generated to minimize the inter-class discrepancy and maximize intra-class discrimination. Applied to the content-based retrieval task, the proposed method can be combined with Euclidean distance for the comparison of low-dimensional features. As demonstrated in the extensive experiments on the benchmark datasets, the performance of the proposed framework is superior to the state-of-the-art methods.

Recent network research has demonstrated that the performance of convolutional neural networks can be improved by introducing a learning block that captures spatial correlations. In this paper, we propose a novel multiple feature reweight DenseNet (MFR-DenseNet) architecture. The MFR-DenseNet improves the representation power of the DenseNet by adaptively recalibrating the channel-wise feature responses and explicitly modeling the interdependencies between the features of different convolutional layers. First, in order to perform dynamic channel-wise feature recalibration, we construct the channel feature reweight DenseNet (CFR-DenseNet) by introducing the squeeze-and-excitation module (SEM) to DenseNet. Then, to model the interdependencies between the features of different convolutional layers, we propose the double squeeze-and-excitation module (DSEM) and construct the inter-layer feature reweight DenseNet (ILFR-DenseNet). In the last step, we designed the MFR-DenseNet by combining the CFR-DenseNet and the ILFR-DenseNet with an ensemble learning approach. Our experiments demonstrate the effectiveness of CFR-DenseNet, ILFR-DenseNet, and MFR-DenseNet. More importantly, the MFR-DenseNet drops the error rate on CIFAR-10 and CIFAR-100 by a large margin with significantly fewer parameters. Our 100-layer MFR-DenseNet (with 7.1M parameters) model achieves competitive results on CIFAR-10 and CIFAR-100 data sets, with test errors of 3.57% and 18.27% respectively, achieving a 4.5% relative improvement on CIFAR-10 and a 5.09% relative improvement on CIFAR-100 over the best result of DenseNet (with 27.2M parameters).

This paper investigates a finite-time formation control problem for multiple networked quadrotors. A novel distributed control approach is presented under the leader-follower formation framework, and the approach is developed based on the finite-time Lyapunov theory and the homogeneous system theory such that all quadrotors form and maintain a desired geometric pattern within finite time while tracking a reference trajectory. The designed control law is composed of a dynamic observer, a position controller and an attitude controller, in which the observer is adopted to provide estimates of the leader quadrotor information for each follower quadrotor, and the controllers are in a cascade structure. It is shown that the finite-time leader-follower formation of a group of quadrotors can be achieved via the distributed control approach, and the cascade control architecture conforms to quadrotor dynamic characteristics. The constructive procedures on how to synthesize such a control law are also given. The effectiveness of the proposed control approach is verified by the simulation.

By introducing a feedback path in the conventional droop control loop of islanded microgrid, the voltage amplitude and frequency restoration can be effectively achieved by using the secondary control. However, the usage of low bandwidth communication (LBC) links, the communication delays and the underlying data drop, would increase the system complexity and reduce its reliability. In our previous studies, it is found that the voltage and frequency deviation of islanded three-phase microgrid in synchronous reference frame can be restored by utilizing a band-pass filter (BPF)-based improved droop control strategy, without any communication lines and additional control loops. For more general scenarios, the BPF-based droop control scheme is first extended to islanded single-phase microgrid in hybrid frame in this paper to achieve deviation restoration. Moreover, the dynamic stability of the studied system is addressed by the derived reduced-order small-signal model in this paper, which simplify the modeling process and theoretical analysis. Followed by the system model, the impact of system parameters variation on the stability and dynamic performance of the microgrid are subsequently predicted by applying the eigenvalue-based analysis approach. Consequently, the analytical results show an overdamped system with good stability and robustness against system parameters drift. Finally, the effectiveness of eigenvalue analysis is verified by simulation results, and the experiment results are provided to further validate the feasibility of BPF-based droop control method in islanded single-phase microgrid.

Resource sharing among vehicles can highly improve the capability and efficiency of Internet of Vehicles (IoV). However, it is challenging to establish trust and preserve privacy during the resource sharing process because of the high mobility and topological variability in IoV. Emerging blockchain technology expresses the excellent performance in handling distributed trust due to its verifiable and immutable ledger. In this paper, we first propose a consortium blockchain-based resource sharing paradigm in IoV, in which the resource sharing interactions are encapsulated as transactions and recorded by Road Side Units (RSUs). Moreover, a lightweight consensus mechanism named as Proof-of-Reputation is proposed to reduce computational power consumption and motivate vehicles involved in resource sharing. Finally a differentiated resource pricing scheme is proposed based on the dynamic match game of resource demand and supply. The reputation value is designed to indicate the trustworthy degree of vehicles, and the trust is established via the consensus procedure. We couple the resource sharing process and consensus together by utilizing the reputation value of each vehicle. The security and privacy analysis as well as simulation experiments on communication performance can verify the efficiency of the proposed blockchain system.

Deep learning has made significant progress in many fields such as image identification, speech recognition and natural language processing, especially in the field of computer vision. The better performance of the neural network often built on deeper, wider network structure, more network parameters and more storage and often computational expensive. As a result, it is hard to deploy neural network to mobile and embedded devices. Therefore, compressing of convolutional neural networks is very necessary and practical. In this paper, we propose a channel pruning algorithm for depth-wise separable convolution units and introduce a new channel selection algorithm based on information gain and a method for quickly recovering network performance after pruning. The proposed method is implemented on MobileNet and validated on several popular datasets. The experimental results show that our method can achieve better experimental results on several image classification datasets, and also achieve good detection results on the PASCAL VOC image detection dataset.

GNSS is now being speedily expanded to our daily life, but the positioning precision still can hardly meet the demands of many high-precision applications, such as approaching landing system on airports. Due to the development of GNSS, quadruple-frequency signals are now available in China's BeiDou Navigation Satellite System (BDS) and the European Galileo system, which can contribute to positioning precision. Positioning precision can not be improved obviously by quadruple-frequency carrier phases until cycle slips are detected and repaired. A method using four linear combinations to detect and repair quadruple-frequency cycle slips is proposed in the paper. The choices of the four linear combinations are conducted in cascaded steps in accordance to the cycle slip fixing probability. When the four detection combinations are determined, cycle slips on original carrier phase observations can be uniquely determined. The proposed algorithm has been tested on real 30-second quadruple-frequency static observations of BDS and Galileo and on real 0.05-second quadruple-frequency kinematic observations of BDS and Galileo. Simulated and real cycle slips are tested. The results show that the proposed algorithm can detect and repair cycle slips even for one cycle effectively.

Iterative learning scheme is proposed with the aim to achieve perfect tracking of a prescribed reference trajectory for systems that operate repetitively. In this paper, a sensor fault estimation framework is proposed for nonlinear repetitive system. First, the problem of sensor fault estimation is converted to actuator fault estimation via state redefinition. Afterward, state observer is designed for state reconstruction while iterative learning law is presented for fault estimation. Thus, the information in the previous period can be utilized to improve the fault estimating performance in current iterative trial. Avoiding the uncertainty caused by the norm optimal theory, the uniform convergence of error extended system is guaranteed by asymptotic stability and optimal function. Finally, the efficiency and merits of the proposed scheme are illustrated by numerical examples.

In order to solve the problem of “abandoned” wind caused by short circuit faults in a wind farm, a wind farm fault locating method based on redundancy parameter estimation is proposed. Using the characteristics of the traveling wave, transmission equations containing the position of the fault point are constructed. Parameter estimation from statistical theory is used to solve the redundant transmission equations formed by multiple measuring points to locate the faults. In addition, the bad data error detection capability of the parameter estimation is used to determine bad data and remove them. This improves locating accuracy. A length coefficient is introduced to solve the error enlargement problem caused by a transmission line sag. The proposed fault locating method can solve the fault branch misjudgment problem caused by the short circuit faults near the data measuring nodes of the wind farm based on the proposed fault interval criterion. It also avoids the requirements to the traveling wave speed of traditional methods, thus its fault location is more accurate. Its effectiveness is verified through simulations in PSCAD/EMTDC, and the results shows that it can be used in the fault locating of hybrid transmission lines.

DENCLUE is a typical density-based clustering method, and it also is an important pattern classification analysis technique. In this clustering method, the propriety of parameters’ values greatly influences the quality of distinguishing. Accordingly, how to choose an appropriate value of every parameter is a problem that is worth studying. The problem is focused in this paper, and a method which is very different from previous ones is presented. The highlight of the method is that the selection of parameters no longer depends on personal experience but on data original distribution. More specifically, smoothing parameter $h$ is more or less proportional to the average value of distance between two arbitrary data points; step size $\delta $ is adapted according to the density of data points in the hill-climbing progress; noise threshold $\xi $ is replaced by $\delta $ and total number of data points. Compared with the original DENCLUE and an improved DENCLUE proposed in our previous research, the optimized algorithm can bring better clustering from experiments. In addition, due to the adaptiveness of $\delta $ , the method will become less complex.

The (InxGa1_x)2O3 photodetectors were fabricated on the single-crystalline (InxGa1_x)2O3 films deposited on sapphire substrate by pulsed laser deposition. The structural and optical properties of the epilayers were investigated using high-resolution X-ray diffraction, X-ray photoelectron spectroscopy, spectroscopic ellipsometry, and transmittance spectra. With decreasing the growth temperature, the indium composition increased and the bandgap decreased from 4.99 eV to 4.89 eV (In0.05Ga0.95)2O3 and 4.78 eV (In0.08Ga0.92)2O3. Furthermore, the photoelectrical characteristics of (InxGa1-x)2O3 detectors were also studied. The enhanced ' photo, ' dark, and responsivity R were achieved in the devices with higher In composition, while a larger number of defects were introduced, resulting in the significant persistent photoconductivity.

Recent studies show that convolutional neural networks (CNNs) has made a series of breakthroughs in the two tasks of face detection and pose estimation, respectively. There are two CNN frameworks for solving these two integrated tasks simultaneously. One is to use face detection network to detect faces firstly, and then use pose estimation network to estimate each face’s pose; the other is to use region proposal algorithm to generate many candidate regions that may contain faces, and then use a single deep multi-task CNN to process these regions for simultaneous face detection and pose estimation. The former’s problem is pose estimation’s performance is affected by face detection network because two networks are separate. The latter generates lots of candidate regions, which will bring huge computation cost to CNN and can’t achieve real-time. To solve the above existing problems, we propose a multi-task CNN cascade framework that integrates these two tasks. We show that multi-task learning of face detection and head pose estimation helps to extract more representative features. We exploit CNN feature fusion strategy to further improve head pose estimation’s performance. We evaluate face detection on FDDB benchmark, and evaluate pose estimation on AFW benchmark. Our method achieves comparative result compared with state-of-the-art in these two tasks and can achieve real-time performance.

In this paper, we report the design and fabrication of fine-pitch and pixelated-addressed micro-light emitting diode (LED) arrays with emission wavelengths of red (R), green (G), blue (B), and infrared (IR). The arrays have a resolution of 8×8 with monochromatic LED chips directly bonded to custom-designed printed circuit boards to form pixelated matrix. R, G, B, and IR micro-LED arrays with a pixel pitch of 1 mm were demonstrated. Sequential row scanning and multiple columns programming is performed by two CMOS-based application specific integrated circuit chips providing constant-sinking current and forward voltage for each pixel of the micro-LED arrays. The uniformity of operating current and forward voltage of the micro-LED pixels was measured with 1.94% coefficient of variation. The micro-LED pixel has much shorter rising time and higher response frequency than those of regular packaged blue and white LEDs. The implemented R-G-B micro-LED arrays are the excellent candidate for high-definition fine-pitch LED displays and demonstrated its versatile functionalities such as visible light communications. The IR micro-LED arrays show great potential in health medical treatment, IR detection imaging and local positioning system.

Small heterogeneous cells have been introduced to improve the system capacity and provide the ubiquitous service requirements. In order to make flexible deployment and management of massive small cells, the utilization of self-powered small cell base stations with energy harvesting (EH-SCBSs) is becoming a promising solution due to low-cost expenditure. However, the deployment of static EH-SCBSs entails several intractable challenges in terms of the randomness of renewable energy arrival and dynamics of traffic load with spatio-temporal fluctuation. To tackle these challenges, we develop a tractable framework of the location deployment and mobility management of EH-SCBSs with various traffic load distributions an environmental energy models. In this paper, the joint optimization problem for location deployment and mobile management is investigated for maximizing the total system utility of both users and network operators. Since the formulated problem is a NP-hard problem, we propose a low-complex algorithm that decouples the joint optimization into the location updating approach and the association matching approach. A suboptimal solution for the optimization problem can be guaranteed using the iteration of two stage approaches. Performance evaluation shows that the proposed schemes can efficiently solve the target problems while striking a better overall system utility, compared with other traditional deployment and management strategies.

Automatically predicting age group and gender from face images acquired in unconstrained conditions is an important and challenging task in many real-world applications. Nevertheless, the conventional methods with manually-designed features on in-the-wild benchmarks are unsatisfactory because of incompetency to tackle large variations in unconstrained images. This difficulty is alleviated to some degree through convolutional neural networks (CNN) for its powerful feature representation. In this paper, we propose a new CNN-based method for age group and gender estimation leveraging residual networks of residual networks (RoR), which exhibits better optimization ability for age group and gender classification than other CNN architectures. Moreover, two modest mechanisms based on observation of the characteristics of age group are presented to further improve the performance of age estimation. In order to further improve the performance and alleviate over-fitting problem, RoR model is pre-trained on ImageNet first, and then it is fune-tuned on the IMDB-WIKI-101 data set for further learning the features of face images, finally, it is used to fine-tune on Adience data set. Our experiments illustrate the effectiveness of RoR method for age and gender estimation in the wild, where it achieves better performance than other CNN methods. Finally, the RoR-152+IMDB-WIKI-101 with two mechanisms achieves new state-of-the-art results on Adience benchmark.

In this paper we analyze the bounds on the reliability of RaptorQ codes under maximum likelihood decoding, especially in the finite-length regime. RaptorQ code ensembles by a high-order low density generator-matrix (LDGM) code as pre-code and an inner Luby transform (LT) code. By investigating the rank of the product of two random coefficient matrices of the high-order LDGM code and the inner LT code, we derive the expressions for the upper and lower bounds of decoding failure probability (DFP) on the RaptorQ code. Then, the accuracy of our derived theoretical bounds are verified through the Monte Carlo simulations with different degree distributions and short packets. The high accuracy bounds are then exploited to design near-optimum finite-length RaptorQ codes, enabling a tight control on the tradeoff between decoding complexity and DFP.

Insulator detection using an infrared image is challenged by variance of temperature, orientations, and a cluttered background. A robust and discriminative representation of insulators in electric power systems is needed. This paper proposes a novel method for generating this type of representation in infrared images by taking advantage of high-level discriminative Convolutional Neural Networks (CNNs) to feature the extraction framework and the deformation invariant nature of the Vector of Locally Aggregated Descriptors (VLAD) aggregator. Different from existing methods, we delve deep into the convolutional feature maps. We first extract deep activation maps from convolutional layers of a pretrained deep model and replace the last three fully-connected layers with a VLAD pooling layer to generate the representation of an insulator. Then, we train a Support Vector Machine (SVM) for binary classification. To further verify the effectiveness and robustness of our proposed feature, an insulator detection pipeline based on an object proposal is introduced. The experimental results show that our proposed method can achieve an accuracy of 93%. Meanwhile, the detection results demonstrate that our insulator detection pipeline has satisfied performance goals.

The smart grid is widely considered as an efficient and intelligent power system. With the aid of communication technologies, the smart grid can enhance the efficiency and reliability of the grid system through intelligent energy management. However, with the development of new energy sources, storage and transmission technologies together with the heterogeneous architecture of the grid network, several new features have been incorporated into the smart grid. These features make the energy trading more complex and pose a significant challenge on designing efficient trading schemes. Based on this motivation, in this paper, we present a comprehensive review of several typical economic incentive approaches adopted in the energy-trading control mechanisms. We focus on the technologies that address the challenges specific to the new features of the smart grid. Furthermore, we investigate the energy trading in a new cloud-based vehicle-to-vehicle energy exchange scenario. We propose an optimal contract-based electricity trading scheme, which efficiently increases the generated profit.

Mobile edge computing (MEC) is a promising paradigm to provide cloud-computing capabilities in close proximity to mobile devices in fifth-generation (5G) networks. In this paper, we study energy-efficient computation offloading (EECO) mechanisms for MEC in 5G heterogeneous networks. We formulate an optimization problem to minimize the energy consumption of the offloading system, where the energy cost of both task computing and file transmission are taken into consideration. Incorporating the multi-access characteristics of the 5G heterogeneous network, we then design an EECO scheme, which jointly optimizes offloading and radio resource allocation to obtain the minimal energy consumption under the latency constraints. Numerical results demonstrate energy efficiency improvement of our proposed EECO scheme.