Search

A batch variable learning rate gradient descent algorithm is proposed to efficiently train a neuro-fuzzy network of zero-order Takagi-Sugeno inference systems. By using the advantages of regularization, the smoothing L1/2 regularization is utilized to find more appropriate sparse network. Combining the second-order information of the smoothing error function, a variable learning rate is chosen along the steep descent direction, which avoids line search procedure and may reduce the cost of computation. In order to appropriately adjust the Lipschitz constant of the smoothing error function in the learning rate, a new scheme is proposed by introducing a hyper-parameter. Also the article applies the modified secant equation for estimating the Lipschitz constant, which makes the algorithm greatly reduce the oscillating phenomenon and improve the robustness. Under appropriate assumptions, a convergent result of the proposed algorithm is also given. Simulation results for two identification and classification problems show that the proposed algorithm has better numerical performance and promotes the sparsity capability of the network, compared with the common batch gradient descent algorithm and a variable learning rate gradient-based algorithm.

Recently, how to exploit unlabeled data for training face recognition models has been attracting increasing attention. However, few works consider the unlabeled shallow data1 in real-world scenarios. The existing semi-supervised face recognition methods that focus on generating pseudo labels or minimizing softmax classification probabilities of the unlabeled data do not work very well on the unlabeled shallow data. It is still a challenge on how to effectively utilize the unlabeled shallow face data to improve the performance of face recognition. In this paper, we propose a novel face recognition method, named VirFace, to effectively exploit the unlabeled shallow data for face recognition. VirFace consists of VirClass and VirInstance. Specifically, VirClass enlarges the inter-class distance by injecting the unlabeled data as new identities, while VirInstance produces virtual instances sampled from the learned distribution of each identity to further enlarge the inter-class distance. To the best of our knowledge, we are the first to tackle the problem of unlabeled shallow face data. Extensive experiments have been conducted on both the small- and large-scale datasets, e.g. LFW and IJB-C, etc, demonstrating the superiority of the proposed method.

Achieving ubiquitous ultra-reliable low latency consensus in centralized wireless communication systems can be costly and hard to scale up. The consensus mechanism, which has been widely utilized in distributed systems, can provide fault tolerance to the critical consensus, even though the individual communication link reliability is relatively low. In this article, a widely used consensus mechanism, Raft, is introduced to the Industrial Internet of Things (IIoT) to achieve ultra-reliable and low latency consensus, where the consensus reliability performance in terms of nodes number and link transmission reliability is investigated. We propose a new concept, Reliability Gain, to show the linear relationship between consensus reliability and communication link transmission reliability. We also find that the time latency of consensus is contradictory to consensus reliability. These conclusions can provide guides to deploy Raft protocol in distributed IIoT systems.

A magnetic-levitation (maglev) Wind Yaw System is proposed to make the nacelle yaw for wind direction without friction. Due to the multi-point suspension nonlinear coupling and higher suspension weight, a double closed-loop control strategy for single-point suspension control is proposed including suspension air gap loop and the current tracking loop. A MDOF dynamic model based on the mechanical geometry and motion dynamics is developed and then transformed into the mult-point suspension model. Then we unify the cross coupling terms and pitching disturbance into synchronous disturbance, and then proposed a composite control strategy to set suspension current reference for current tracking loop, including main tracking controller, adaptive synchronization controller and adaptive disturbance controller. The multi-point suspension simulation platform of maglev wind yawing system is built to carry out the dynamic suspension tracking and pitching disturbance experiment, and the poposed strategy has good dynamic tracking ability, the steady-state air gap error is only 0.003mm, and the reference switching recovery time is 0.445s . As pitching torque of 220Nm is subjected to the nacelle, the maximum air gap fluctuation is only 0.38mm, and the maximum synchronization error is 0.19mm, but these of the PID controller can reach to 2.82mm and 10mm respectively, which validate the proposed composite control and the rationality of four point nacelle suspension.

Signal coordination methods have been widely investigated to improve the performance of the traffic control system in urban arterial. The traditional solution is centralized, which has high computational complexity. To solve the problem, this paper constructs a linear model by assuming that the cycle of each intersection is known. And we proposed a parallel optimization method for arterial signal coordination, which uses the framework of the ADMM algorithm. Consistency constraints are introduced to decompose the centralized problem into multiple independent sub-problems, and parallel solving is realized. Compared with linear programming, the precision of ADMM-based algorithm is 99.5% in the multi-intersection coordination control tasks. Besides, with the increase of arterial road scale, the computational complexity of this method increases linearly, which has the potential to improve the efficiency of large-scale signal coordination.

The splitting strategy over prediction horizon of model predictive control (MPC) has the potential to compute optimal action in a parallel way. However, such time-splitting algorithms often lead to very slow convergence speed because the state consensus only happens in each pair of adjacent nodes, i.e. a point-to-point topology. This paper proposes a generic cross-node consensus method to extend the shortcoming of limiting to point-to-point topology for the purpose of accelerating the convergence of time-splitting MPC. The cross-node consensus is realized by predicting the state transition from one node to another using plant prediction model, which can increase the information exchange efficiency in the prediction horizon. The time-splitting optimization algorithm is implemented by combing with alternating directions method of multipliers (ADMM). Simulations with autonomous driving show that this new algorithm significantly reduces the number of iterations in time-splitting MPC, averagely about 81% compared with classic time-splitting technique.

As one of the common diseases, hollowing deterioration destroys the structure and shortens the life of stone relics. The hollowing deterioration samples for the stone relics were nondestructive tested and studied by terahertz time-domain spectroscopy (THz-TDS) in this paper. The hollowing samples with different parameters can be distinguished by the THz spectra of the samples, which imply that THz nondestructive testing (THz NDT) method is applicable to test hollowing deterioration for stone relics. The THz NDT method not only can provide reliable reference data for repair and maintenance of stone relics, but also is an economic nondestructive detection technique with promising application prospects.

Noise produced by electrical equipment in the UHV converter station during operation may adversely affect the surrounding environment. In order to identify the contribution of equipment noise in the station from the total noise at the station boundary, it is necessary to conduct an in-depth study on the spectrum characteristics and identification methods of main noise sources in the converter station, so as to control the noise of the equipment in the station in a targeted manner. In this paper, basic characteristics of the source noise spectrum were analyzed based on the measured spectral data around the source equipment. Then, reasonable timbre parameters were selected. Finally, spectral characteristics of main noise sources equipment in the UHV converter station were compared based on the timbre parameters, and characteristic timbre parameters of the main sound sources equipment in the station are obtained. According to these characteristic timbre parameters, the identification method of main noise sources equipment noise in the converter station were given, and the noise source in multiple converter stations is identified by this method. Results show that the noise source noise identification method’s the accuracy of UHV converter station based on timbre parameters is more than 85%, which proves the accuracy of the method. The identification method proposed in this paper can be applied to study the noise of noise sources equipment inside the station of converter station.

In many real-world face restoration applications, e.g., smartphone photo albums and old films, multiple high-quality (HQ) images of the same person usually are available for a given degraded low-quality (LQ) observation. However, most existing guided face restoration methods are based on single HQ exemplar image, and are limited in properly exploiting guidance for improving the generalization ability to unknown degradation process. To address these issues, this paper suggests to enhance blind face restoration performance by utilizing multi-exemplar images and adaptive fusion of features from guidance and degraded images. First, given a degraded observation, we select the optimal guidance based on the weighted affine distance on landmark sets, where the landmark weights are learned to make the guidance image optimized to HQ image reconstruction. Second, moving least-square and adaptive instance normalization are leveraged for {spatial} alignment and illumination translation of guidance image in the feature space. Finally, for better feature fusion, multiple adaptive spatial feature fusion (ASFF) layers are introduced to incorporate guidance features in an adaptive and progressive manner, resulting in our ASFFNet. Experiments show that our ASFFNet performs favorably in terms of quantitative and qualitative evaluation, and is effective in generating photo-realistic results on real-world LQ images. The source code and models are available at https://github.com/csxmli2016/ASFFNet.

Symmetric positive definite matrix has a widely application in many branches of science and engineering, a key issue is about how to find the ‘nice’ distance between two elements on $SPD(n)$ consisting of symmetric positive definite matrices. By using the fiber bundle and constructing a Riemannian submersion from the general linear group $GL(n)$ to $SPD(n)$ , we develop the closed-form expressions of a new Riemannian metric and associated Riemannian distance in $SPD(n)$ . This metric and distance can be used to obtain a more precise result compared with the widely used Euclidean distance, which is usually used in signal classification.

Environmental sound classification (ESC) is a significant problem in signal processing and machine learning. In this paper, a novel unsupervised learning approach is proposed using topological data analysis. Each sound signal is transformed into a point cloud by time-delay embedding. The persistence diagrams are obtained after computing the persistent homology of Vietoris-Rips filtrations. The sound signals are assigned to different classes using kmeans algorithm based on the bottleneck distance matrix of persistence diagrams.

With the rapid development of augmented reality technology in recent years, the application of mobile AR has become more and more extensive. This paper introduces an AR mobile application based on ARcore, which aims to popularize AR life style, let users experience the fun of AR, combine AR functions such as plane detection, multi image recognition with interesting real scenes, and create new AR playing methods such as transmission door and magic box. The application is divided into four parts, the first part and the third part are presented in a virtual way, the second part and the fourth part are presented in a combination of virtual and reality. Players constantly switch between the virtual world and the real world to experience the ultimate AR fun.

Other than classroom work, recommending learning resource in the extra-curricular activities is also an important factor to improve students' disciplinary study environment. And the learning resource, such as extra-curricular book, supporting video and other learning materials, recommending in extra-curricular activities requires the matching of the requirements of the student and the disciplinary of the resource. Taking the disciplinary book as an example, this paper firstly analyzes the problem of disciplinary-requirement mismatch between the student and the disciplinary book in extra-curricular activities, and describes the student's requirement for disciplinary book recommendation system. Based on requirement investigation, the specific demands for book recommendation system based on android are analyzed. Finally, the disciplinary book recommendation system based on android for disciplinary study in extra-curricular activities is designed.

An adaptive aggregate smoothing gradient type algorithm is proposed for training fuzzy neural networks with max-min fuzzy logic operator. The study in this paper extends the work in [Long Li et al. A convergent smoothing algorithm for training max-min fuzzy neural networks. Neurocomputing (2017), https://doi.org/10.1016/j.neucom.2017.04.046]. An adaptive precision parameter adjustment rule is applied to manage the increasing precision parameter. It is shown in numerical experiment that the proposed algorithm can avoid the oscillation phenomenon and improve learning capacity effectively.

The optical transmission characteristics for the hollow carbon-coated Fe3O4 colloidal photonic crystal have been calculated with the finite-difference time-domain (FDTD) method. We analyze the influence of the factors on the photonic band gap (PBG) that include lattice constant a, the number of the particles in propagating direction Ny, the thickness of carbon layer Hc and Fe3O4 cluster layer Hf, and the thickness ratio of the two layers. The results show that the PBGs red shift and the bandwidth first increases and then decreases with the increasing a. In the situation of increasing Ny, the PBG changes from irregular to uniform, followed by the oscillations on both sides of the PBG growing in number and the deepened PBG in the low-frequency region. The PBGs move toward the low frequency direction with the increase of Hc, and the optimal value of Hc for the uniform color response is 10 nm ~25 nm. The PBGs red shift with the increasing Hf, and the first bandwidth increases while the second decreases. The optimal Hf for the ideal PBG is 35 nm ~55 nm. The stop bands move to the high-frequency direction with the increasing thickness ratio (Hc : Hf), and the best ratio is 10 nm : 55 nm for the complete PBG and wider bandwidth.

Cloud robotic system needs to learn real time data stream in order to adapt to the new environment for better performance. Incremental learning as a kind of dynamics classification method can avoid sample repeating learning, and develop model for updating itself. In this paper, we proposed an incremental learning-based cloud robot object recognition mechanism. It can learn new objects through human-robot interaction by speech recognition. The proposed incremental learning method was tested with small scale data in domestic indoor environment application, and the cloud robotic system was validated by basic conversation with different testing groups. The experimental results verify the feasibility of the incrementally learning new information algorithm and the proposed cloud robotic system.

Coordination control problem of networked Euler-Lagrange systems with external disturbance is addressed. The sliding mode surface and coordination control law of networked Euler-Lagrange systems with a dynamic leader under a directed graph is first proposed. It is shown that the followers can track the dynamic leader in finite time. Then leaderless consensus law for networked Euler-Lagrange systems is proposed. Based on algebraic graph theory and Lyapunov stability theory, finite-time stability of the closed-loop system is proven. The simulation results show the feasibility and effectiveness of the proposed control law.

Aging population and disabled assistance have become serious social problems these years. However, it is impossible to make sure that every old people has a caregiver. Therefore, the home service robot is a good method to mitigate these issues. Although current robotic technology is mature, service robots with advanced functions are very expensive. In this paper, we developed an identity recognition low-cost home service robot based on Turtlebot and Robot Operating System (ROS). In order to verify the performance of the service robot, we designed a series of experiments in a simulated home environment. The average successful rates of the designed experiments were all above 90%. The results confirmed that the low-cost service robot satisfied our design target.

The health care is in increasingly high demand with the aging population. In this paper, we present a novel health care platform for elderly people and patients who need to be taken care. The platform consists of a walk-aid unit, medical apparatus for health monitoring, data storage unit and remote monitoring unit. The medical devices are used for the elderly or patients who are undergoing a rehabilitative treatment, to determine the body's vital physiological indexes like pulse rate, heart rate, blood pressure level, body temperature and the inertial navigation system (INS). The inertial navigation system (INS) is employed to capture people's abnormal action like tumble. The walk-aid platform is designed to enhance people's mobility and collect position information of the elderly in case of an emergency. It is also equipped with a camera to capture a picture of the elderly and the environment. The data storage unit can collect and store the data from the medical devices and the walk-aid unit. The remote monitoring unit like pad or smart phone can be connected with the platform to check the situation of users. And in turn, if an emergency situation happened, the server computer can also send a message to the mobile device. The case study shows that it is an efficient way to keep the user under observation.

This paper is to design attitude coordination control strategies for multiple spacecraft by state feedback and output feedback. Based on Lyapunov theory and graph theory, first, bounded coordination synchronization controllers are developed by state feedback. Second, output feedback controllers are developed for coordination synchronization(i.e. without angular velocity feedback), it can guarantee spacecraft attitude reach to stable attitude. Finally, simulation results illustrate and visualize the effectiveness and feasibility of the proposed control technique.

Modular multilevel converters (MMCs) become more and more attractive in high-voltage and high-power applications. The balance of the floating capacitor voltages is vital for MMCs. In order to balance the capacitor voltages, a fundamental frequency sorting strategy with Phase Disposition PWM is proposed in this paper. A dual sorting method, which sorts the capacitor voltage increments and the present capacitor voltages simultaneously, is implemented to change the relationship between the carriers and the sub-modules once in every fundamental period. Due to the low sorting frequency, the intensive computation of the controllers is avoided. Moreover, the switching commutations are evenly distributed among the power devices. Furthermore, the measurement of the arm current is eliminated, which further simplifies the communication among different controllers. Finally, the proposed voltage balance strategy is validated by simulation and experiment.

This paper proposes verification for the feasibility of human-robot coexistence and interaction simulation in SIGVerse environment. SIGVerse is a human-robot interaction environment, which can be also applied for virtual reality. It solves the problem for lacking of human-robot interaction in simulation software previously. In this paper, a robot following human task in the domestic indoor environment is presented to test the software. In the experiment, binocular visual servo planar mobile robot is used for completing two confusion challenges, binocular vision distance measurement and extended PD controller are used to complete the task, which are set to test the universal of the algorithm and the feasibility of the simulation. Then the measuring and following algorithm is tested in the real environment as well. The experimental result shows that it is possible for same algorithm using for both simulation and real environment.

In LTE-A (Long Term Evolution-Advanced) network deployed with higher order MIMO (Multiple Input Multiple Output), CRS (Common Reference Signal) is substituted by CSI-RS (Channel State Information Reference Signal) aiming for detecting channel more efficiently. But LTE (Long Term Evolution) legacy UEs (User Equipment) cannot distinguish CSI-RS from data which may deteriorate the LTE UEs' demodulation performance. And in CoMP (Coordinated Multiple Point Transmission) scenario the number of REs (Resource Element) occupied by CSI-RS will be multiplied, in which case the demodulation performance would be even worse. So the introduction of CSI-RS makes it troublesome for LTE legacy UEs scheduled in the LTE-A network. Simulation results by both link-level and system-level reveal that LTE legacy UEs' performance is more seriously degraded due to the impact of CSI-RS when the coding rate and modulation order go higher. In this paper we propose and compare different MCS (Modulation Coding Scheme) selection algorithms about reducing the impact of CSI-RS on LTE legacy UEs scheduled in LTE-A network. SLA (Smart Link Adaptation) algorithm relys on manned control too much for different scenarios. IE-SINR (Interpolation Empirical SINR) algorithm is applicable to different scenarios while it cannot improve system performance as much as SLA algorithm. VSE (Valid Spectrum Efficiency) algorithm is a more general method and can adapt to different scenarios and simulation results show that VSE algorithm can improve system throughput not lower than SLA algorithm.

Cell Range Expansion (CRE) technology provide an opportunity for more users to transfer from macro cell to small cell in the Long Term Evolution Advanced (LTE-A) heterogeneous networks. However, these CRE users located at the edge of small cells will be seriously interfered by macro cell, resulting in a sharp decline to their Signal to Interference plus Noise Ratio (SINR). In order to reduce this interference, a method of frequency domain self-organizing interference coordination technique combined with CRE technology is proposed in our paper. In this method, the total spectrum bandwidth is divided into two parts according to the frequency partition ratio which can be adaptively controlled. One part of the bandwidth in which the transmission power of macro cell is reduced. Another band with the transmission power of macro cell do not change. Based on this method, small cell users with low SINR including CRE users and macro cell users with high SINR scheduled in the power reduced band. Our simulation results show that the proposed scheme in this paper can effectively improve the system's throughput and increase the small cell's edge users' performance.

In LTE-A (Long Term Evolution-Advanced) network, CSI-RS (Channel State Information Reference Signal) is substituted for CRS (Common Reference signal) to perform higher order MIMO (Multiple Input Multiple Output). In CoMP (Coordinated Multiple Point Transmission) scenario, the number of REs (Resource Element) occupied by CSI-RS would be multiplied. However, LTE (Long Term Evolution) legacy UEs (User Equipment) cannot distinguish CSI-RS from data, which may deteriorate the LTE UE's demodulation performance. Therefore it could be troublesome for LTE legacy UEs scheduled in LTE-A network with higher order MIMO equipped, and it is even more serious when CoMP technology is used. Investigation results by both link-level and system-level simulation reveal that LTE legacy UEs performance is seriously degraded due to the impact of CSI-RS when coding rate and modulation order go higher. To reduce the performance degeneration, an improved scheduling algorithm named AUP-AMCS (Advanced UE Prioritization combined with Adaptive Modulation Coding Scheme) is proposed in this paper. Compared to existing algorithm, AUP-AMCS algorithm takes UE priority into consideration and is compatible with the established specification. Simulation results indicate that the proposed algorithm can improve system performance obviously.

In the Long Term Evolution Advanced (LTE-A) heterogeneous networks (Het-Nets), smallcell is an emerging and key item, providing improved service coverage and high spectral efficiency with low power. However, the number and location arrangement of smallcell is always changing in reality, the interference scenario is also different. In order to reduce the layer-to-layer interference in uncertain deployment of smallcell, a method of frequency domain self-organizing interference coordination technology is proposed in this paper. In this method, the total spectrum bandwidth is divided into two parts according to a frequency partition ratio. One part of the bandwidth is called protected band where macrocell's transmission power will be reduced in an adaptively controlled way to schedule the edge UEs of smallcell and the center UEs of macrocell, another is named unprotected band to schedule the macro edge UEs and smallcells' center UEs. Thus, the edge UEs of smallcell could be protected under the condition of not losing too much macrocells' performance. The system-level simulation results show that the proposed scheme can achieve superior system performance improvements, and higher cell-edge user data rate.

The passenger car logistics transportation loading problem of company H is investigated. Based on reasonable assumptions, the passenger car logistics transportation loading problem is transformed into a two-dimensional packing problem. Firstly, under the premise of considering the sizes of two types of car carriers and three kinds of passenger cars, and the stability requirements of car carriers, 36 kinds of possible loading modes of type 1-1 car carriers and 73 kinds of possible loading modes of type 1-2 car carriers are enumerated by using MATLAB program. Then, the integer programming model of passenger car logistics transportation loading problem is constructed to minimize the total transportation cost. A LINGO program is given to solve the model. Using the mathematical model and LINGO program, we obtain the number of two types of car carriers and the loading mode of each car carrier which company H can use to complete the given transport task. The results of this paper can provide a theoretical basis for the logistics company to make passenger car logistics transportation plans.

Nowadays, Voice of LTE (VoLTE) is used to support voice communications in LTE network. As one of the key techniques for VoLTE, semi-persistent scheduling (SPS) is important to improve the capacity of VoLTE users. However with SPS, the resource and modulation and coding scheme (MCS) are usually decided by the initial scheduling and keep unchanged during the active state. Considering the continuously changing channel, it is hard to ensure the performance of the voice packet. In order to solve this problem, a resource allocation algorithm is proposed based on the prediction of channel quality change. Furthermore, the data rate of Adaptive multi Rate(AMR) codec is adaptively adjusted in the resource allocation algorithm to reduce VoLTE's impact on other data services. Simulation results prove that the proposed algorithm can reduce the retransmission of VoLTE users and improve the throughput performance of other traffic businesses.