Search

Phase-change materials (PCMs) have become as promising elements in silicon photonic systems. We propose an optical buffer device based on vanadium dioxide (VO2) embedded on silicon-on-insulator waveguides with a microring resonator. The upload straight waveguide is for coupling light signal from the input port into VO2 embedded on microring resonator and the download straight waveguide with VO2 films is for coupling the storage signal to the output port. The optical characteristics of optical buffer device under different structural parameters are analyzed by using finite difference mode simulation (MODE). By controlling the phase state of VO2 in two coupling regions, the buffer speed of the device can be up to 0.5 ps and buffer time up to 78.28 ps. The optical buffer based on VO2 is a new way for PCM optical buffer.

With the rapid urbanization during the past three decades in China, the urban expressway network has arisen with large numbers in many cities. However, the traffic conditions are becoming worse and worse. One of the main reasons is that the connected sections of main road and expressway do not have a coordination by the intersection signal and variable speed limits (VSL) on expressway. In this paper, a new cooperation optimization model of VSL on urban expressway and the intersection signal timing are proposed based on the mesoscopic traffic flow model. Its objective is to minimize the total delay and the number of parking. Its essence is adjusting traffic flow on the network in space-time so that the whole network can be in a state of equilibrium. It is effective to the control of local traffic congestion and can avoid the spread of congestion, thus improving traffic safety. Additionally, the optimal VSL strategy is proposed. And then a solution algorithm based on the simultaneous perturbation stochastic approximation (SPSA) is designed for the proposed model. In the end, a practical example of Yanshan overpass in Ji’nan elevated expressway is used to validate the effectiveness of the proposed method. The simulation results are satisfactory. The conclusions are very useful for traffic management department to implement reasonable traffic control strategy.

L-shaped array (LA) is traditionally used as two one-dimensional uniform linear arrays (ULAs) for two-dimensional direction of arrival (DOA) estimations. With elevation and azimuth angles estimated separately, the detectable source number is strictly limited by the number of elements in the ULAs. To increase the detection ability, in this paper, we propose to exploit the diff-sum co-array (DSCA) concept on the L-shaped array. As such, a virtual element in the quadrants of DSCA can be generated by only one pair of sensors in the LA. Therefore, it can acquire much higher number of degrees of freedom for DOA estimation. To further reduce the redundancy of elements on the horizontal and vertical axes of DSCA, an element moving strategy is proposed. It is proved that the DSCA can remain intact with any one or more physical sensors relocated to the centrosymmetric positions. Based on this strategy, a generalized L-shape array concept and a new configuration referred to as generalized L-shaped array with odd-even locations (GLA-OEL) are developed. For GLA-OEL, the number of sensor pairs with small spacing is as small as one, which dramatically reduces the mutual coupling of adjacent elements. Besides, with the same number of sensors, the GLA-OEL can generate much larger virtual uniform rectangular array in the DSCA. Simulations verify the superiorities of the proposed array in terms of detection performance and estimation accuracy.

The trajectory planning time is uncertain when there is no velocity constraint in the joint space of an operating robot. This paper proposes a 3-5-3 polynomial interpolation trajectory planning algorithm based on improved cuckoo search algorithm (ICS) which functions under a velocity constraint. Kinematics analysis of the robot is conducted and the 3-5-3 polynomial interpolation function is established as the foundation for trajectory planning. The size of the step size control factor is set as a variant which varies with the number of iterations, which accelerates the algorithm’s convergence and prevents premature falling into the local optimal solution, both problems in the traditional cuckoo search algorithm where the step size control factor is fixed. The proposed algorithm was used to optimize the trajectory of a test robot with interpolation time as the search space for optimization. Any interpolation time that did not meet the speed constraint was eliminated to secure the shortest possible robot running time. The UR robot was taken as a research object to simulate the joint motion trajectory planning in MATLAB. The proposed algorithm performed well and better realized the time-optimal trajectory under the constraint of speed compared to the traditional cuckoo search algorithm, particle swarm optimization algorithm, and genetic algorithm.

To overcome the shortcomings of conventional methods in solving inverse kinematics problems of redundant degree-of-freedom robots, this paper converts the inverse kinematics problems of the manipulator into target optimization problems, and it presents an Improved quantum particle swarm optimization algorithm (IQPSO), which is used to solve the inverse kinematics problems. In this algorithm, quantum behavior is added to the particle swarm optimization algorithm, and the improved contraction expansion coefficient with first large and then small is adopted, which can not only traverse the whole search space, but also improve the convergence speed and solution accuracy. Based on the forward kinematics equation, this paper takes the position error of the end-effector of the robot and the minimum energy consumption in the process of the robot motion as the optimization objectives, and it conducts simulation experiments on a 7-degree-of-freedom (7-DOF) robot. The experimental result shows that the IQPSO algorithm has faster convergence speed and higher solution accuracy than the traditional particle swarm optimization algorithm (PSO) and quantum particle swarm optimization algorithm (QPSO). It is an effective method to solve the inverse kinematics problem of the robot.

Knowledge distillation aims to transfer representation ability from a teacher model to a student model. Previous approaches focus on either individual representation distillation or inter-sample similarity preservation. While we argue that the inter-sample relation conveys abundant information and needs to be distilled in a more effective way. In this paper, we propose a novel knowledge distillation method, namely Complementary Relation Contrastive Distillation (CRCD), to transfer the structural knowledge from the teacher to the student. Specifically, we estimate the mutual relation in an anchor-based way and distill the anchor-student relation under the supervision of its corresponding anchor-teacher relation. To make it more robust, mutual relations are modeled by two complementary elements: the feature and its gradient. Furthermore, the low bound of mutual information between the anchor-teacher relation distribution and the anchor-student relation distribution is maximized via relation contrastive loss, which can distill both the sample representation and the inter-sample relations. Experiments on different benchmarks demonstrate the effectiveness of our proposed CRCD., CVPR2021 Poster

C 5 F 10 O as an environment-friendly SF6 substitute gas attracts a lot of attention due to its excellent insulation performance and low global warming potential (GWP). However, further researches on insulation and arc extinguishing performance of C 5 F 10 O are needed before large-scale practical application. In this paper, insulation characteristic test platform was established to test arc extinguishing performance through arc interruption, and insulation performance through lighting impulse (LI), and creeping discharge. It was found that the positive and negative lightning impulse voltages of C 5 F 10 O/CO 2 gas mixture with a partial pressure of 40 kPa at 0.7 MPa were 276.3 kV and 161.46 kV, which were respectively 91.20% and 88.10% of that of 0.4 MPa SF 6 . The average successful and unsuccessful breaking current of the C 5 F 10 O/CO 2 gas mixture with a partial pressure of 20 kPa at 0.7 MPa were 1 730.2 A and 1 967.2 A, which were 89.7% and 91.7% of that of 0.4 MPa SF6. The results show that C 5 F 10 O/CO 2 gas mixture is a promising alternative to SF6 in terms of insulation and arc extinguishing performance.

C5F10O gas insulating medium has attracted wide attention in the power industry for its excellent environmental potential and insulation performance. When C5F10O gas is applied in power industry engineering, it will inevitably face the over thermal fault which often occurs inside the equipment. Under the long-term effect of the local over thermal fault, whether C5F10O itself can maintain stability is an important factor to determine its engineering application. However, the thermal stability characteristics of C5F10O and its mixture are still unknown. Therefore, through systematic experimental research on the existing experimental platform of partial over thermal of the gas insulating medium, this paper explores the initial decomposition temperature of 5% C5F10O and 95% CO2 mixture, the formation conditions of decomposition components, the relationship between the generation characteristics of decomposition components and fault temperature, and so on. The results show that when the pressure of 5% C5F10O and 95% CO2 is 0.2MPa, the mixture begins to decompose at about 400° to form C3F8 and C3F6. C2F6 and CF4 will be produced at above 500°. The process of C3F6 formation is the endothermic reaction, while the process of C3F8, C2F6 and CF4 formation is an exothermic reaction, which is very beneficial to the formation of C3F6 un-der superheated conditions. As a result, the product concentration shows as C3F6 > C3F8 > C2F6 > CF4, and the rate of product formation is positively correlated with the over thermal fault temperature. The results of this study lay a foundation for the application of C5F10O gas in the electric industry.

This paper studies the secure transmission in cognitive wiretap relaying networks with full-duplex technique over Rayleigh channels, where a secondary user (SU) transmitter (Alice) aims to transmit confidential information to a multi-antenna legitimate SU receiver (Bob) in the presence of a multi-antenna relay (Relay), a primary user (PU) and a passive eavesdropper (Eve). For secrecy performance enhancement, we propose maximal-ratio combining-zero forcing beamforming/maximal-ratio transmission-selection combining-zero forcing beamforming (MRC-ZFB/MRT-SC-ZFB) scheme. The closed-form expression for the secrecy outage probability (SOP) of the CRNs with MRC-ZFB/MRT-SC-ZFB scheme is derived. From the analysis, we look into the high signal-to-noise ratio (SNR) regime, and present simple secrecy outage approximation for the proposed scheme, which reveals that the secrecy diversity order is min(N R , N R N B ) , where N R and N B are the number of antennas at Relay and Bob, respectively.

In recent years, deep neural networks have been widely concerned by researchers in facial expression recognition. However, insufficient facial training data of the public available database is a major challenge in deep learning, which will lead to an obvious decrease in the effectiveness of learning result; many data augmentation techniques have thus been widely used to enrich the training dataset. In this paper, we introduce a contextual loss function to construct a Contextual Generation Adversarial Network with one generator and one discriminator. The proposed method can map the neutral expression to six basic expressions to expand the database. The experimental results on CK + and KDEF databases show that the proposed method can effectively improve the ability to extract facial features and the ability to generate higher quality images. The data augmentation used the proposed method improves the recognition rate of facial expressions on KDEF and CK + datasets.

Remote photoplethysmography (rPPG) using a simple consumer-level webcam has great potential for healthcare, human-computer interaction, affective computing, and many other fields. However, traditional methods based on the skin reflection model and signal processing achieved limited success. Thus, this paper presents a novel end-to-end spatial-temporal convolutional neural network for the rPPG recovering task: To preserve the details, the pooling layer is moved to the branch network, and the attention mechanism is used to enhance the backbone network performance. Moreover, the ConvLSTM was used to learn the relationship between frames. The manual features extraction and signal post-processing steps can be omitted through our end-to-end neural network. The experimental results conducted on PURE and COHFACE datasets show that the proposed method significantly improved the quality of the recovered rPPG signal.

Due to the cross-sensitivity to kinds of gases, traditional single sensor is impossible to selectively detect gases, which limits its application. In this paper, a sensor array with supervised and unsupervised algorithms was employed to selectively detect NO 2 , CO and their mixtures. To improve the recognition accuracy, average resistance over a period of time was introduced to acquire features, including response value, response time and recovery time. Firstly, principal component analysis (PCA) was utilized to reduce the dimensionality of samples. With the help of unsupervised fuzzy C-means clustering algorithm, samples have exhibited obvious clustering. Furthermore, artificial neural network (ANN) has reached a high accuracy of quantitative identification on six different detected gases varying from 0 to 50ppm.

A novel method of point cloud data processing is proposed in this paper. This method is dedicated to the low-speed self-driving cars based on a lidar. The purpose of point cloud data processing is to reduce the amount of computation and its storage space of the 3D raw point cloud. Meanwhile, the processed point cloud is different from the ordinary 2D point cloud and it does not lose key spatial information points. The procedure is divided into three steps. Firstly, a valid data field is selected. And then the selected data field is filtered by the statistical outlier removal algorithm. Secondly, the point cloud filtered is downsampled by the voxel grid filter. Finally, the downsampled 3D point cloud data are processed to 2D data regularly. The experimental results show that the proposed method can retain the key spatial information points of the raw point cloud. The processed point cloud requires less computing resources and storage space, have higher reliability compared with the ordinary 2D data, reduce the algorithm complexity compared with 3D data, and be able to real time processing. The experimental vehicle equipped with a velodyne-16 sensor which outputs 3D point cloud data. As a comparison of the experiments, the vehicle also contains an rplidar_A2 sensor which outputs ordinary 2D data.

In this paper, a dielectric resonator antenna (DRA) with self-diplexing property is proposed for the first time. Based on the high flexibility of a dielectric resonator, two modes of a rectangular dielectric resonator (TE δ11 and TE δ13 ) are utilized to implement two independent operating modes of an antenna. By introducing two pairs of open stubs on the feed lines, good isolation between two input ports is achieved together with improved bandpass characteristics. For demonstration, a self-diplexing DRA operating at 1.8 GHz and 2.45 GHz was designed, fabricated and measured. Reasonable agreement between simulation and measurement can be observed. The diplexing function has been successfully implemented without introducing extra filters. In addition, similar radiation properties have been realized for the two operating modes.

When the single-phase PV (photovoltaic) grid-connected inverter with H6 topology applies conventional PI controller to track the sinusoidal reference current, the steady-state-error will be generated and its anti-interference capability becomes poor. To deal with this problem, in this paper an compound current control strategy based on quasi-PR (Proportional Resonant) and repetitive control is proposed. The fundamental of the compound current controller is introduced, and the tuning method of the controller parameters is analyzed. The grid-connected current harmonic suppression performance of the designed compound current is validated on a single-phase photovoltaic grid-connected inverter experiment platform.

During the operation of power grid companies for many years, due to various technical reasons, non-standard management or power and responsibility adjustment, the end-user meter files do not conform to the actual situation, and it is difficult to determine the location of the ground cable power users. Further, the disordered accounts cause that the analysis of line loss rate, distribution network fault location, emergency repair work list issuance and three-phase imbalance analysis and some other advanced applications cannot be effectively carried out. Seriously affected the process of building smart grid by Power Grid Corp. For this reason, this paper proposes a transformer area recognition method based on improved k-Means clustering algorithm. Based on the high correlation of voltage data of the same electrical energy acquisition equipment, the voltage data of users are clustered to realize the users in the same transformer area. Examples show that the proposed transformer area recognition method can accurately and effectively identify the user transformer area and phase information and reduce the labor and hardware costs. It is of great significance to guide the operation, maintenance, emergency repair, technical renovation and planning of low-voltage distribution network

Traditional methods for fault diagnosis of transmitters are mainly based on expert systems. This method requires manual extraction of data features, which is inevitably affected by human factors, and is prone to diagnostic errors. Aimed at this problem, this paper proposes deep learning theory to extract the deep feature representation of fault information, and construct a fault diagnosis model based on Stacked Auto-Encoder. Firstly, the collected fault signals are normalized, and the network is trained layer by layer using classless data. Then the network parameters are fine-tuned according to the tagged fault data, the fault signal features are automatically extracted. Finally, the trained network is used to detect various fault signals. Through simulation analysis, this method effectively reduces the diagnostic error, and the diagnostic performance is superior to the traditional diagnostic methods.

intelligent vehicles, especially autonomous ones, have been studied in recent years. This paper proposes an intelligent vehicle control system solution. An upper-computer is employed to process sensor information detecting road conditions, while a STM32F103 microprocessor is used to implement vehicle motion control. The upper computer and the lower-MCU communicate with each other through the Modbus/RTU protocol. All the software in the designed is open source, scalable and at low cost. Experimental results show that the control system is feasible and effective.

this paper is based on the replacement of Sulfur hexafluoride (SF 6 ) gas in switchgears, by using C 4 F 7 N; a fluorinated nitrile. The principle research is however required to examine the potential of fluorinated Nitrile as compared to the SF 6 . Until now, lesser number of attempts being made to investigate the arc interruption time by usingC 4 F 7 N. The objective of this work was to investigate the arcing time ofC 4 F 7 N with different mixture compositions and compare the results with SF 6 . C 4 F 7 N due to limitations liquefaction temperature, cannot be use alone, so this paper will show the mixture of (C 4 F 7 N) withCO 2 as the main subject. By mixing a small amount ofC 4 F 7 N intoCO 2 , the breaking performance of the insulating gas can be significantly improved. Under the conditions, that the partial pressure ofC 4 F 7 N suitable replacement of SF6, C 4 F 7 N is certain, increasing the total pressure of the mixed gas ofC 4 F 7 N/CO 2 can also improve its breaking performance

SF 6 has been widely used as an insulating gas, which can decompose into several gaseous byproducts in electric discharges, like arc, spark or corona (partial) discharge, in the presence of the impurities H 2 O and O 2 . Researching on the decomposition products is helpful for improving the personal safety and long-time reliability of SF 6 insulated equipment. The decomposition characteristics of SF 6 under corona discharge is not only related to the insulation fault type but also has close relationship with the concentration of H 2 O in the insulated equipment. The effects of H 2 O on the decomposition products and possible main reactions have been analyzed a lot, however, the influence mechanism remains unaddressed. This work is devoted to obtaining the effects of the concentration of H 2 O on the decomposition products and the main reaction pathways. In this paper, the chemical kinetic model is established for 50Hz ac point-plane corona discharge. The multi-step time is used to obtain the rapid discharge process in the plasma region and the slow reaction process in the gas region simultaneously. The concentrations of SOF 2 , SO 2 F 2 and SO 2 and the main pathways of these three species under different moisture content at 12h are obtained. The moisture content varies from 200ppm to 5000ppm with the oxygen content keeping at 100ppm.

This paper introduces a localization method where GPS, inertial measurement unit (IMU), odometer and 3D lidar are the main sensors. GPS is mainly used as a global position, and is also used as an initial position for map matching. The map matching is realized through the Normal Distributions transform (NDT)point cloud matching algorithm, by which the optimal transformation of the lidar point cloud and the established map is obtained. In case of GPS signal loss, the dead reckoning which is realized by the fusion of IMU and odometer is also applied. By comparing the experimental results to the accurate traj ectory, it is found that the positioning accuracy has been improved using the algorithm herein.

Chinese sentence similarity calculation is widely used in the field of natural language processing, and the similarity calculation method is also varied. This paper introduces the similarity calculation method based on shallow Information, based on the similarity of part of speech, and also analyzes their calculation principles and calculation methods. Their advantages and disadvantages are given. At the same time, it puts forward the idea of using deep learning thinking to automatically learn the essential information of data from large-scale text data. Using the word2vec model, sentence processing is reduced to word vectors. It also proposes a method for calculating the similarity of shallow features, parts of speech features, and features of word vectors. The experimental results show that compared with the conventional method, the accuracy of the algorithm in calculating the sentence similarity is higher.

Safety and reliability are the basic requirements of integrated energy systems (IES). It is of great significance to carry out risk assessment research to ensure the reliability of energy supply of IES. In this paper, an electrical-thermal energy flow model with steam as the heat medium and its calculation algorithm are proposed firstly. Then the IES risk assessment hierarchy considering interests of both the system operator and users is designed, and the IES risk assessment procedure based on electrical-thermal energy flow is proposed. The results of the test case show that the proposed assessment method can effectively identify the overall risk of the multienergy flow system.