Search

Existing studies for the balance control of unmanned bicycle robots only consider constant forward velocity and a single optimal objective that cannot be applied to the complex motion situation. To balance the bicycle robot with time‐varying forward velocity, only with the steering actuator, the multiple objective optimal balance control issue is studied here. A fuzzy state‐space model under different forward velocities is firstly offered based on the non‐linear Euler–Lagrange model. Based on this, a closed‐loop equation under a fuzzy controller is verified. To regulate the feedback gain of the fuzzy controller, a modified particle swarm optimization (MPSO) algorithm with two stages is proposed. In the MPSO's second stage, a novel objective fitness function, consisting of multiple objectives and combining the conventional Hurwitz stability analysis criterium, is designed. Procedures for the MPSO dynamic programming approach are presented. By two examples, the efficiency of the MPSO algorithm, for time‐varying and time‐constant velocity situations, and faster capacity for iteration convergence, are examined.

Channel estimation plays a significant role in the Inter-carrier interference (ISI) mitigation and symbol detection in Orthogonal frequency division multiplexing (OFDM) systems. As the speed of mobile terminal increases, the channel changes rapidly, causing fast fading which will degrade the system performance. In this paper, a low complexity fast fading channel estimation method suited for scattered pilot OFDM systems is proposed. This method is characterized by estimating channel matrix directly in the frequency domain, which can significantly reduce the channel estimation complexity with good system performance. Simulation results verify its superiority over the existing estimation methods.

The cable-driven segmented manipulator with cable-constrained synchronous rotating mechanisms (CCSRM) is a new type of continuum manipulator, which has large stiffness and less motors, and thus exhibits excellent comprehensive performance. This paper presents a dynamic modeling method for this type of manipulator to analyze the friction and deformation of the cables on the dynamical behaviors of the system. First, the driving cables are modeled based on the ALE formulation, the strategies for detecting stick-slip transitions are proposed by using a trial-and-error algorithm, and the stiff problems of the dynamic equations are released by a model smoothing method. Second, the dynamic modeling method for rigid links is presented by using quaternion parameters. Third, the connecting cables are modeled by torsional spring-dampers and the frictions between the connecting cables and the conduits are considered based on a modified Coulomb friction model. Finally, the numerical results are presented and verified by comparing with experiment results. The study shows that the friction and cable deformation play an important role in the dynamical behaviors of the manipulator. Due to these two factors, the constant curvature bending of the segments does not remain.

Human visual system relies on both monocular focusness cues and binocular stereo cues to gain effective 3D perception. Correspondingly, depth from focus/defocus (DfF/DfD) and stereo matching are two most studied passive depth sensing schemes, which are traditionally solved in separate tracks. However, the two techniques are essentially complementary: the monocular cue from DfF/DfD can robustly handle repetitive textures and occlusion that are problematic for stereo matching whereas the binocular cue from stereo matching is insensitive to defocus blurs and can resolve large depth range. In this paper, we emulate human perception and present unified learning-based techniques to conduct hybrid DfF/DfD and stereo matching. We first construct a comprehensive focal stack dataset synthesized by depth-guided light field rendering. Next, we propose different network architectures to suit various inputs, including focal stack, stereo image pair, binocular focal stack, a focus-defocus image pair and defocus-stereo image triplet. We also exploit different connection methods between the separate networks for integrating them into an optimized solution to produce high fidelity disparity maps. For experiment, we further explore different hardware setup to capture both monocular and binocular depth cues. Results show that our new learning-based hybrid techniques can significantly improve accuracy and robustness in depth estimation.

Limited by the operation time window and working space, space teleoperation tasks need to be completed within an expected time while ensuring that the end effector meets the physical constraints. Meanwhile, the interaction with unknown environments would cause uncertainty in the closed-loop system, which brings great challenges to the control design. To solve the above problems, the control performance issue for a class of space teleoperation systems subject to multiple constraints and interaction uncertainties is investigated in this article. The force interaction with the human operator/space environment is represented by interval type-2 (IT2) Takagi-Sugeno (T–S) fuzzy systems, where the uncertain equivalent mass and damping parameters can be effectively described and captured by IT2 membership functions. In order to reduce the communication burden and satisfy the constraints of settling time, transient-state performance and operating space, a time-varying threshold event-triggered control scheme together with exponential-type Lyapunov function is developed for the first time. We show that, with the proposed controller, the synchronization tracking errors are guaranteed to converge to a user-defined residual set within preassigned settling time, and never exceed the prescribed range despite unknown control direction and actuator faults, which solves the long-standing constraint issue with more flexibility due to the fact that the related constraints can be arbitrarily specific within the physically available range. Moreover, the convergence set is only dependent on fewer user-defined parameters rather than approximation errors, which provides an effective analysis technique to deal with the difficulty that the convergence accuracy is difficult to calculate quantitatively in the presence of unknown disturbance. Detailed simulation results are provided to show the effectiveness and merit of the proposed control strategy.

Single-track two-wheeled robots have become an important research topic in recent years, owing to their simple structure, energy savings and ability to run on narrow roads. However, the ramp jump remains a challenging task. In this study, we propose to realize a single-track two-wheeled robot ramp jump. We present a control method that employs continuous action reinforcement learning techniques for single-track two-wheeled robot control. We design a novel reward function for reinforcement learning, optimize the dimensions of the action space, and enable training under the deep deterministic policy gradient algorithm. Finally, we validate the control method through simulation experiments and successfully realize the single-track two-wheeled robot ramp jump task. Simulation results validate that the control method is effective and has several advantages over high-dimension action space control, reinforcement learning control of sparse reward function and discrete action reinforcement learning control.

Accurate and reliable fault diagnosis is one of the key and difficult issues in mechanical condition monitoring. In recent years, Convolutional Neural Network (CNN) has been widely used in mechanical condition monitoring, which is also a great breakthrough in the field of bearing fault diagnosis. However, CNN can only extract local features of signals. The model accuracy and generalization of the original vibration signals are very low in the process of vibration signal processing only by CNN. Based on the above problems, this paper improves the traditional convolution layer of CNN, and builds the learning module (local feature learning block, LFLB) of the local characteristics. At the same time, the Long Short-Term Memory (LSTM) is introduced into the network, which is used to extract the global features. This paper proposes the new neural network—improved CNN-LSTM network. The extracted deep feature is used for fault classification. The improved CNN-LSTM network is applied to the processing of the vibration signal of the faulty bearing collected by the bearing failure laboratory of Inner Mongolia University of science and technology. The results show that the accuracy of the improved CNN-LSTM network on the same batch test set is 98.75%, which is about 24% higher than that of the traditional CNN. The proposed network is applied to the bearing data collection of Western Reserve University under the condition that the network parameters remain unchanged. The experiment shows that the improved CNN-LSTM network has better generalization than the traditional CNN.

Accurate segmentation of the prostate on magnetic resonance (MR) images plays an important role for prostate cancer diagnosis and treatment. Although many automated prostate segmentation methods have been proposed, the performance still faces several challenges, which includes large variability in prostate shape, unclear boundary, and complex intensity distribution. Therefore, the results obtained from the automated methods should be further refined by users to get a more accurate and reliable segmentation. In this paper, we propose an end-to-end interactive segmentation method to refine the automated results. A convolutional long short term memory (convLSTM) module and a gated graph neural network (GGNN) are presented in the proposed method for prostate segmentation in both automated and interactive manners. A boundary loss is proposed to train our model. We evaluated the proposed method on two public available datasets and one in–house dataset. Experimental results show that the proposed convLSTM module could obtain a DSC of 91.78% on the test dataset, which outperforms eight state-of-the-art methods. A further 1.5% improvements can be obtained by user interactions based on the GGNN. The segmentation time including user interactions and inference time was 2.3 min on average for segmenting one volume.

This letter studies the problem of air-to-air visual detection of micro unmanned aerial vehicles (UAVs) by monocular cameras. This problem is important for many applications such as vision-based swarming of UAVs, malicious UAV detection, and see-and-avoid systems for UAVs. Although deep learning methods have exhibited superior performance in many object detection tasks, their potential for UAV detection has not been well explored. As the first main contribution of this letter, we present a new dataset, named Det-Fly, which consists of more than 13 000 images of a flying target UAV acquired by another flying UAV. Compared to the existing datasets, the proposed one is more comprehensive in the sense that it covers a wide range of practical scenarios with different background scenes, viewing angles, relative distance, flying altitude, and lightning conditions. The second main contribution of this letter is to present an experimental evaluation of eight representative deep-learning algorithms based on the proposed dataset. To the best of our knowledge, this is the first comprehensive experimental evaluation of deep learning algorithms for the task of visual UAV detection so far. The evaluation results highlight some key challenges in the problem of air-to-air UAV detection and suggest potential ways to develop new algorithms in the future. The dataset is available at https://github.com/Jake-WU/Det-Fly .

Nature Products Research (NPR) has a long history of revealing bioactive constituents of natural origin, both as single drug leads within modern western medicine and as mixtures of bioactive constituents enriching traditional medicines. Identifying bioactive constituents in complex mixtures such as those obtained from extracting marine algae has been relying on multidisciplinary techniques, such as bioactivity-guided or spectroscopic-guided fractionation and purification. In this regard, milestones of scientific achievements of NPR have been hailed by applying novel technologies, such as improved separation or purification, spectroscopic hardware with detection limits of natural abundance, software algorithms for accelerating data collecting and processing, and high-throughput screening.In most NPR, the characterization of novel compounds as well as the dereplication of known compounds entails the collection and analysis of NMR spectra. This involves the running of 1D and 2D NMR spectroscopic experiments for the purpose of partial structure construction, assemblage and relative stereochemistry determination. As exciting advancements in the rapid genetic and proteomic approaches have made their way into NPR, conventional NMR practices have become one of several bottlenecks in the characterization and dereplication of new compounds. In regard to this challenge, we leveraged the advantages of Non Uniform Sampling Nuclear Magnetic Resonance (NUS NMR) and Artificial Intelligence (AI) to create Small Molecule Accurate Recognition Technology (SMART) as a tool to speed up marine natural products discovery. Fast NMR techniques like NUS NMR have the potential to further reduce detection limits while maintaining the same sampling time and quality. Next, we applied over 4000 experimental Heteronuclear Single Quantum Correlation (HSQC) spectra for the AI training. The outcome is that the AI algorithm provided us with structurally insightful AI embedding maps with nodes and clusters representing correlations of related families of natural products. By testing different HSQC spectra using this algorithm, we can greatly accelerate the rate of known compound identification as well as rapidly generating hypotheses about the relationship of new molecules to those used for the training - based entirely on their NMR properties. Specifically, the 2D NMR spectra of a series of unknown compounds isolated from two different marine cyanobacteria were recognized by the SMART belonging to a specific class of marine depsipeptides.

In this article, Hopfield neural networks system with time-varying delays driven by nonlinear colored noise is introduced. The existence and globally exponential stability of stationary solutions are investigated for such random delay neural networks systems, which may be regarded as a generalization for the case of the constant equilibrium point in the literature. Moreover, the synchronization behavior of linearly coupled delay Hopfield neural networks driven by nonlinear colored noise is investigated at the level of the random attractor. Finally, illustrative examples and numerical simulations are provided to show the effectiveness of the obtained results.

Using objective vibration evaluation to produce results highly consistent with real road ride comfort is challenging. The deficiencies in traditional evaluation indices, adopting an average operator, maximum operator, or cumulative operator as the main vibration information integration logic, are reported here through 19 designed road field tests in which major vibration information distribution covers all axes and vibration information is distributed in spacetime in various patterns. A new evaluation index which adopted a combination of maximum and cumulative operator is proposed to overcome these deficiencies and an interactive mechanism which standardized the process of selecting vibration information distributed among axes and spacetime is devised between the localized major vibrations. The results show that the proposed road ride comfort evaluation index is more consistent and accurate than the evaluation indices proposed by ISO 2631-1 and can be used more generally.

In this article, a difference-coarray-based direction of arrival (DOA) method is introduced, which utilizes the uniform linear array (ULA) in a novel fashion to address the problem of DOA estimation for coherent signals. Inspired by the coarray-based estimators employed in cases of sparse arrays, we convert the sample covariance matrix of the observed signals into the difference coarray domain and process the signals using a spatial smoothing technique. The proposed method exhibits good accuracy and robustness in both the uncorrelated and coherent cases. Numerical simulations verify that the ULA difference coarray- (UDC-) based method can achieve good DOA estimation accuracy even when the SNR is very low. In addition, the UDC-based method is insensitive to the number of snapshots. Under extremely challenging conditions, the proposed UDC-ES-DOA method is preferred because of its outstanding robustness, while the UDC-MUSIC method is suitable for most moderate cases of lower complexity. Due to its demonstrated advantages, the proposed method is a promising and competitive solution for practical DOA estimation, especially for low-SNR or snapshot-limited applications.

Tactile representation on touchscreens plays an important role in improving realism and richness of users' interaction experience. The dynamic lateral force range and the efficient feedback dimensions are very critical in determining the fidelity of tactile displays. This article develops a tri-modal Electrovibration, Ultrasonic Vibration, and Mechanical Vibration (EUMV) tactile display integrating three types of representative principles, which enhances the dynamic lateral force range by leveraging electrostatic and ultrasonic vibrations stimuli, and induces the normal feedback dimension by utilizing mechanical vibration stimulus. Then, a tactile perception scheme with the EUMV display is proposed for simultaneously rendering contour and texture roughness features of visualized surfaces, in which the contour gradient-lateral force model and the texture gradient-perceived roughness model are determined respectively. Objective and subjective evaluations with 20 participants show that the novel scheme establishes significant improvements in both correct recognition ratios of geometric shapes and tactile perception realism of visualized images than the previous studies.

At present, the tactile perception of 3D geometric bumps (such as sinusoidal bumps, Gaussian bumps, triangular bumps, etc.) on touchscreens is mainly realized by mapping the local gradients of rendered virtual surfaces to lateral electrostatic friction, while maintaining the constant normal feedback force. The latest study has shown that the recognition rate of 3D visual objects with electrovibration is lower by 27$\%$ than that using force-feedback devices. Based on the custom-designed tactile display coupling with electrovibration and mechanical vibration stimuli, this paper proposes a novel tactile rendering algorithm of 3D geometric bumps, which simultaneously generates the lateral and the normal perceptual dimensions. Specifically, a mapping relationship with the electrostatic friction proportional to the gradient of 3D geometric bumps is firstly established. Then, resorting to the angle between the lateral friction force and the normal feedback force, a rendering model of the normal feedback force using mechanical vibration is further determined. Compared to the previous works with electrovibration, objective evaluations with 12 participants showed that the novel version significantly improved recognition rates of 3D bumps on touchscreens.

Purpose Accurate and robust segmentation of the prostate from magnetic resonance (MR) images is extensively applied in many clinical applications in prostate cancer diagnosis and treatment. The purpose of this study is the development of a robust interactive segmentation method for accurate segmentation of the prostate from MR images. Methods We propose an interactive segmentation method based on a graph convolutional network (GCN) to refine the automatically segmented results. An atrous multiscale convolutional neural network (CNN) encoder is proposed to learn representative features to obtain accurate segmentations. Based on the multiscale feature, a GCN block is presented to predict the prostate contour in both automatic and interactive manners. To preserve the prostate boundary details and effectively train the GCN, a contour matching loss is proposed. The performance of the proposed algorithm was evaluated on 41 in-house MR subjects and 30 PROMISE12 test subjects. Result The proposed method yields mean Dice similarity coefficients of 93.8 ± 1.2% and 94.4 ± 1.0% on our in-house and PROMISE12 datasets, respectively. The experimental results show that the proposed method outperforms several state-of-the-art segmentation methods. Conclusion The proposed interactive segmentation method based on the GCN can accurately segment the prostate from MR images. Our method has a variety of applications in prostate cancer imaging.

This paper is concerned with Hopfield neural networks with unbounded time-varying delay driven by G-Brownian motion. The existence and uniqueness of solutions, as well as the continuity of solution...

Image segmentation technology has made a remarkable effect in medical image analysis and processing, which is used to help physicians get a more accurate diagnosis. Manual segmentation of the medical image requires a lot of effort by professionals, which is also a subjective task. Therefore, developing an advanced segmentation method is an essential demand. We propose an end-to-end segmentation method for medical images, which mimics physicians delineating a region of interest (ROI) on the medical image in a multi-step manner. This multi-step operation improves the performance from a coarse result to a fine result progressively. In this paper, the segmentation process is formulated as a Markov decision process and solved by a deep reinforcement learning (DRL) algorithm, which trains an agent for segmenting ROI in images. The agent performs a serial action to delineate the ROI. We define the action as a set of continuous parameters. Then, we adopted a DRL algorithm called deep deterministic policy gradient to learn the segmentation model in continuous action space. The experimental result shows that the proposed method has 7.24% improved to the state-of-the-art method on three prostate MR data sets and has 3.52% improved on one retinal fundus image data set.

With the trend of outsourcing fabrication, split manufacturing is regarded as a promising way to both acquire the high-end nodes in untrusted external foundries and protect the design from potential attackers. However, in this article, we show that split manufacturing is not inherently secure, that a hardware Trojan attacker can still recover necessary information with a proximity-based or a simulated-annealing-based mapping approach together with a probability-based or net-based pruning method at the placement level. We further propose a defense approach by moving the insecure gates away from their easily attacked candidate locations. Results on benchmark circuits show the effectiveness of our proposed methods.

It is quite difficult to imitate the motion of human arms using non-humanoid robots due to their dissimilar embodiments (degree-of-freedom, body morphology, and constraints). However, in most cases of the robotic imitation, the human operator and the robot would not share the same kinematic configuration. This paper addresses the motion imitation problem between the human arm and an industrial robot, where a commonly-used UR5 robot is considered. The motion of the human arm is obtained by an inertial motion capture system, and then the captured motion is reproduced using the UR5 embodiment. A virtual-joint-based approach is proposed to facilitate the fast and intuitive kinematics mapping between the human arm and the UR5 robot, leading to a robotic imitation system that can imitate the tip location and configuration of the human arm simultaneously. The proposed approach is verified experimentally on a real UR5 robot and compared with classic Cartesian-space-based mapping approach and joint-space-based approach.

Interactive image segmentation can improve segmentation performance using manual intervention. Traditional interactive segmentation methods have unsatisfactory segmentation accuracy for images with complex background. Deep learning-based methods depend on large and accurate annotated datasets. In this paper, we propose an online interactive segmentation method based on graph convolutional network (GCN), which includes the superiorities of these two types of methods. We present a pre-segmentation stage to get an initial segmentation of the image, then propose an interactive GCN (iGCN) module to further improve the accuracy of the initial segmentation. Moreover, iGCN module is trained online without any pre-training burden. Experimental results show that our method outperforms several state-of-the-art methods on GrabCut and Berkeley datasets.

Multi-channel speech enhancement is gaining increasing interest in recent years. By combining the beamforming framework with the deep neural network, significant improvement on speech enhancement performance has been achieved. While the neural beamformers designed for the distributed microphone arrays are deployed for practical applications such as teleconferencing and surveillance, there is less approach designed for the co-located microphone arrays such as the soundfield microphones. In this work, a new neural beamforming network is proposed for B-format 3D multi-channel speech enhancement and recognition. The proposed method incorporates the traditional beamforming structure with the deep neural network specifically for the B-format channels (the first-order Ambisonics). The proposed method has ranked the 1st place of the 3D Speech Enhancement task in the MLSP L3DAS21 Challenge while significantly outperformed the baseline system on the WER and STOI metrics.

More and more security situation awareness strategies in power network use deep learning as the data processing method. However, the endogenous security problem of deep learning poses a challenge to the robustness required for power network security situation awareness, and gives attackers an opportunity to add small perturbation to the network and fail the security situation awareness strategy. In this paper, we build the deep neural networks for the power network dataset. We verify a variety of adversarial example attack methods against deep neural networks, and tested the success rate of attacks on different types of networks to compare the difference of attack methods and target model. We can see from the experiment results that adversarial example attacks can have a great impact on the power network security situation awareness and pose a threat to the power network.

Safe driving and stable paths are essential for the autonomous vehicle navigation, that large and fast steering angle should be avoided. This paper addresses a local path generation problem for autonomous vehicles while the least maximum curvature and the shortest length are obtained with limited curvature rate. The properties of the novel feasible paths based on continuous curvature turns are investigated. And a simple and fast computational method is presented to solve the problem by iterative procedure. The simulation in this paper shows that, compared with recent researches, the performance of least maximum curvature and shortest length is obtained by the novel local path planner, and the driving behavior is closer to the human driver operation.

The traditional master-slave teleoperation relies on human expertise without correction mechanisms, resulting in excessive physical and mental workloads. To address these issues, a co-pilot-in-the-loop control framework is investigated for cooperative teleoperation. A deep deterministic policy gradient(DDPG) based agent is realised to effectively restore the master operators' intents without prior knowledge on time delay. The proposed framework allows for introducing an operator (i.e., co-pilot) to generate commands at the slave side, whose weights are optimally assigned online through DDPG-based arbitration, thereby enhancing the command robustness in the case of possible human operational errors. With the help of interval type-2(IT2) Takagi-Sugeno (T-S) fuzzy identification, force feedback can be reconstructed at the master side without a sense of delay, thus ensuring the telepresence performance in the force-sensor-free scenarios. Two experimental applications validate the effectiveness of the proposed framework., 6 pages, 7 figures, accepted by IECON 2021

This work focuses on motion control of the bicycle robot by using the proposed NRRL algorithm. Unlike the traditional RL algorithm, decomposing the main tasks into subtasks manually and introducing qualitative prior knowledge to the agent have been applied in the NRRL algorithm. Simulation results show that better performance and better sample efficiency of the proposed NRRL algorithm have been achieved in terms of balance control and path tracking of bicycle robot. It's believed that the NRRL algorithm is available on the real physical bicycle robot, and the deployment of the algorithm will be realized soon, as the real physical bicycle robot has been constructed currently.

With the development of large-scale power grids, distributed economic dispatch has received considerable attention. However, existing distributed economic dispatch algorithms ignore potential network attacks, which may have a greater impact on the security of the power grid. In addition, due to the access of renewable power generation equipment and external interference, there are usually uncertain terms in the cost function. A distributed economic dispatch collaborative deep reinforcement learning algorithm considering the safety objective function is proposed. All agents in the algorithm make joint decisions by observing the environment and coordinating with local neighbors. The state action value function is approximated by a neural network. Aiming at the problem caused by the uncertain term of the objective function, the idea of lenient reinforcement learning is adopted, and the reward is also fitted with a neural network. Several case studies have been conducted to prove the advantages of this algorithm.

Deep reinforcement learning (DRL) has been proved to be a very promising method for robot motion control. However, it usually needs a large number of samples in training, which restricts its application in real-world robots. Sim-to-Real means transferring training strategies in simulation to reality, and it has become one of the hottest research areas in recent years. Embodiment of DRL algorithms is a necessary step in Sim-to-Real, which is faced with many challenges, such as poor sample efficiency, wear and tear of robots, safety, etc. In this paper, we present a new algorithm called action mapping and state prediction (AMSP), which considers three main factors in training including inaccurate parameters, unmodeled action damping and action delay. This method includes model error compensation based on action mapping, and delay compensation based on state prediction. The method in this paper is demonstrated in OpenAI inverted pendulum environment, and the strategy trained in the ideal environments with no action damping and no action delay is successfully transferred in the form of zero-shot to the artificial simulation environment with action damping and action delay, which shows the effectiveness of AMSP.

The theory, design and implementation of fixed- frequency hysteresis current control of six-phase permanent magnet synchronous motor (PMSM) is presented. Based on the difference between the given current and the actual current value with the loop width, and the algorithm calculates the duty cycle of the turn-on and turn-off states in real-time, The shortcomings of indeterminate switching frequency and large current distortion in traditional hysteresis control was solved by the proposed technique, the fault-tolerant operation of the motor under open- circuit fault can also be guaranteed. Simulation and experimental results are included to show the effectiveness of the proposed current controller.

The bicycle robot is one of the Single Track Two-wheel robot that its balance control is so important as the forward velocity control. The existing researches on balance control focus on the single-objective control, in reality multiple objectives are required to be considered, that they are not proper for the complicate autonomous control. The multi-objective optimization issue for the bicycle robot with constant forward velocity is studied in this paper. The State-space model is verified based on the nonlinear Euler-Lagrange model. A new fitness function with multiple objectives is proposed for the Advanced Particle Swarm Optimization to tune the gains in the feedback controller. The conventional Lyapunov method is involved in the APSO process. The simulation example prove the validity of the given method in this paper.

Most of the current research on Named Entity Recognition (NER) in the Chinese domain is based on the assumption that annotated data are adequate. However, in many scenarios, the sufficient amount of annotated data required for Chinese NER task is difficult to obtain, resulting in poor performance of machine learning methods. In view of this situation, this paper tries to excavate the information contained in the massive unlabeled raw text data and utilize it to enhance the performance of Chinese NER task. A deep learning model combined with Transfer Learning technique is proposed in this paper. This method can be leveraged in some domains where there is a large amount of unlabeled text data and a small amount of annotated data. The experiment results show that the proposed method performs well on different sized datasets, and this method also avoids errors that occur during the word segmentation process. We also evaluate the effect of transfer learning from different aspects through a series of experiments.

In this work, a novel exponential-type Barrier Lyapunov Function (EBLF) is proposed for nonlinear systems with constraint requirements. The most prominent feature of EBLF is that it can be used in a unified scheme, which deals with full state constrained and output constrained problems of SISO/MIMO systems. Together with the EBLF, a novel adaptive velocity reconstruction scheme is presented for a class of space teleoperation systems with external disturbances. Full-state observation errors are guaranteed to converge without violation of the prescribed constraints. Guaranteed constraint observer performance is available. Moreover, the settling time is only dependent on the observer parameters but not initial values or external disturbances. The design and analysis of the observer and the controller are independent. Simulation results further demonstrate the properties of fixed-time and guaranteed-constraint performance of the proposed scheme.

Massive public resume data emerging on the internet indicates individual-related characteristics in terms of profile and career experiences. Resume Analysis (RA) provides opportunities for many applications, such as recruitment trend predict, talent seeking and evaluation. Existing RA studies either largely rely on the knowledge of domain experts, or leverage classic statistical or data mining models to identify and filter explicit attributes based on pre-defined rules. However, they fail to discover the latent semantic information from semi-structured resume text, i.e., individual career progress trajectory and social-relations, which are otherwise vital to comprehensive understanding of people’s career evolving patterns. Besides, when dealing with large numbers of resumes, how to properly visualize such semantic information to reduce the information load and to support better human cognition is also challenging. To tackle these issues, we propose a visual analytics system called ResumeVis to mine and visualize resume data. First, a text mining-based approach is presented to extract semantic information. Then, a set of visualizations are devised to represent the semantic information in multiple perspectives. Through interactive exploration on ResumeVis performed by domain experts, the following tasks can be accomplished: to trace individual career evolving trajectory; to mine latent social-relations among individuals; and to hold the full picture of massive resumes’ collective mobility. Case studies with over 2,500 government officer resumes demonstrate the effectiveness of our system.

Visual privacy concerns associated with image sharing is a critical issue that need to be addressed to enable safe and lawful use of online social platforms. Users of social media platforms often suffer from no guidance in sharing sensitive images in public, and often face with social and legal consequences. Given the recent success of visual attention based deep learning methods in measuring abstract phenomena like image memorability, we are motivated to investigate whether visual attention based methods could be useful in measuring psychophysical phenomena like “privacy sensitivity”. In this paper we propose PrivAttNet - a visual attention based approach, that can be trained end-to-end to estimate the privacy sensitivity of images without explicitly detecting sensitive objects and attributes present in the image. We show that our PrivAttNet model outperforms various SOTA and baseline strategies – a 1.6 fold reduction in $L1$ – error over SOTA and 7%–10% improvement in Spearman-rank correlation between the predicted and ground truth sensitivity scores. Additionally, the attention maps from PrivAttNet are found to be useful in directing the users to the regions that are responsible for generating the privacy risk score.

In modern transportation, pavement is one of the most important civil infrastructures for the movement of vehicles and pedestrians. Pavement service quality and service life are of great importance for civil engineers as they directly affect the regular service for the users. Therefore, monitoring the health status of pavement before irreversible damage occurs is essential for timely maintenance, which in turn ensures public transportation safety. Many pavement damages can be detected and analyzed by monitoring the structure dynamic responses and evaluating road surface conditions. Advanced technologies can be employed for the collection and analysis of such data, including various intrusive sensing techniques, image processing techniques, and machine learning methods. This review summarizes the state-of-the-art of these three technologies in pavement engineering in recent years and suggests possible developments for future pavement monitoring and analysis based on these approaches.

Space debris division of PMO is devoted to play an important role in the research of orbital mechanics and the development of space debris observation techniques in China. To support the work, this division builds and operates a large optical network across China. The observational network include two major subsets, large field survey telescopes and precise tracking telescopes. The main objective of the network is to maintain cataloged orbital debris and detect uncatalogued debris. CHES at YaoAn site is the most recent facility, for higher orbits survey. New operational software and data pipeline that aims at detecting unresolved objects has being developed. This paper presents the details of this facility and related activities.

Object tracking, as one of the open topics in the past few decades, has been widely applied in the field of video processing. Although there has been intensive research on this topic, some challenges still exist, such as occlusions, clutter and dynamic scenarios. In this paper, we propose a new multi-layer formulation for multi-object tracking, which incorporates the traditional optical flow constraints with a new product term. Then, a sequential convex programming (SCP) based method is presented to solve the resulting non-convex optimization problem. The proposed method can achieve a linear convergence rate, which is demonstrated by our numerical results. In order to illustrate the performance of our approach, it is implemented for composite videos, where two objects move in opposite directions. Different experimental settings based on noise-free, occlusions, clutter and dynamic scenarios show that the method can robustly recover multiple layers and the velocity of the objects. Moreover, the experiments indicate considerable potential for handling more complex scenarios.

Aiming at the problem of face image restoration, a super resolution image restoration algorithm based on wavelet and bicubic interpolation is proposed. Low-resolution image is decomposed by wavelet transform. Then, the high frequency sub images of the original low-resolution image are interpolated in the wavelet transform domain, and the high-resolution image is obtained by the bicubic interpolation of the low-resolution image. The experimental results show that the image restoration algorithm based on wavelet and bicubic interpolation is better than the traditional interpolation algorithms (nearest neighbor interpolation, bilinear interpolation, bicubic interpolation) in face image restoration.

Digital dentistry plays a pivotal role in dental healthcare. A critical step in many digital dental systems is to accurately delineate individual teeth and the gingiva in the three-dimension (3D) intraoral scanned (IOS) mesh data. However, previous state-of-the-art methods are either time-consuming or error-prone, hence hinder their clinical applicability. In this paper, we present an accurate, efficient, and fully-automated deep learning model, trained on a dataset of 4,000 IOS data annotated by experienced human experts. On a hold-out dataset of 200 scans, our model achieves a per-face accuracy, average-area accuracy and area under the receiver operating characteristic curve (AUC) of 96.94%, 98.26%, and 0.9991, respectively, significantly outperforming the state-of-the-art baseline. In addition, our model only takes about 24 seconds to generate segmentation outputs, as compared to over 5 minutes by the baseline and 15 minutes by human experts. A clinical performance test of 500 patients with malocclusion or/and abnormal teeth shows that 96.9% of the segmentations are satisfactory for clinical applications, 2.9% automatically trigger alarms for human improvement, and only 0.2% of them need rework. Our research demonstrates the potential for deep learning to improve the efficacy and efficiency of dental treatment and digital dentistry.

Owing to the advantages of high efficiency and high power density, LLC resonant converters have been widely used. Generally, in order to design the closed-loop regulator, the small-signal model is necessary. Nevertheless, the small-signal model of LLC resonant converter is usually complex and inconvenient to be used to design regulators. In this paper, a third-order small-signal model of LLC resonant converter is derived, which is simple enough to obtain analytical expressions and conducive to be utilized to design the closed-up regulator. Finally, the accuracy of the model is confirmed by the simulation results.

In this paper, a new cable-driven segmented manipulator is studied, which maintains a nearly piecewise constant curvature under the constraint of a cable-constrained synchronous rotating mechanism. The main purpose of this paper is to present a refined dynamics model for the cable-driven segmented manipulator system considering friction effects and deformation of the cables. First, the arbitrary-Lagrangian–Eulerian method is introduced for the modeling of the driving cables, and the friction between the cables and the manipulator is described by the LuGre model. Second, the dynamics model and algorithm for the 2D segmented manipulator system considering friction effects and deformation of the cables is established using Lagrange’s equation of the first kind and Baumgarte’s stabilization method. Finally, a single-segment manipulator and a multi-segment manipulator are simulated successively, and an example of the manipulator circumventing an obstacle is simulated. The control errors caused by friction and deformation of the cables are analyzed, and the influences of the joint angle and the cable pre-tightening force on the control errors are discussed. In addition, the control errors of different segments are compared and analyzed, and reduced by adjusting the pre-tightening force of the cables.

The control moment gyroscope (CMG) can provide large roll restoring moment for unmanned motorcycles or bicycles to improve their stability and mobility. However, in most previous CMG-based studies, the balance controllers are on the basis of the inverted pendulum model of the motorcycle or bicycle, i.e., the steering mechanism is locked. As the steering mechanism is a necessary component for the motorcycle and it can also provide a restoring moment for the vehicle, the combined control for balance between CMG and steering is necessary and meaningful. Noting this, this paper studies the dynamic modeling of a motorcycle and proposes a combined balance control method with both steering and double CMGs. A multibody model is presented for dynamic simulation based on Lagrange equations of the first kind. A simplified model is proposed for controller design, based on which a combined balance controller with both steering and double CMGs is developed. The numerical examples are presented to demonstrate the rationality of the simplified model and the effectiveness of the combined balance controller. The simulation results show that the proposed combined controller performs better than the steering controller and the inverted-pendulum-based controller.

The stationary multi-pinhole (MPH) SPECT has the potential to achieve dynamic human myocardial perfusion imaging (MPI) with high sensitivity and high resolution. In this work, the sensitivity of MPH SPECT is 0.06%, four-fold higher than conventional SPECT (typically 0.015%), the dual-head SPECT was set up 77.6 degree. Acquisition time of stationary MPH SPECT was about 5 minutes, and Siemens Symbia T16 took approximately 15 minutes. For stationary MPH SPECT system, dedicated image reconstruction algorithm was developed. System calibration is necessary to ensure installation accuracy and obtain mechanical information for generating system transfer matrix, which will be employed to iterative reconstruction. Projection data were reconstructed using a Pixel-based Ordered Subset-expectation Maximization (POSEM) reconstruction algorithm with prior information. Reconstructed results showed that the image resolution obtained with stationary MPH is significantly better than Siemens Symbia T16, the image quality is also greatly improved compared to rotational LEHR SPECT. Meanwhile, the short acquisition time of MPH can reduce the possibility of artifacts due to patient movement. So it is potentially capable of dynamic 3D acquisition, and further more clinical trials for evaluating the system performances will be conducted.

Extensive research results have shown that animals like pigeons and turtles can use geomagnetic information for long-distance migration and homing. This article studies the bionic navigation method inspired by magnetotaxis behavior without prior knowledge. The problem of bionic geomagnetic navigation is generalized as an autonomous search of navigation path under the excitation of geomagnetic environment. The geomagnetic gradient-assisted evolutionary algorithm for long-range underwater navigation is proposed. In order to optimize the navigation path, the heading angle predicted by the geomagnetic gradient is used to constrain the sample space in the evolutionary algorithm. Then, according to the principle of multiparameter simultaneous convergence, the evaluation function is improved to enhance the reliability and accuracy of the navigation path. Simulations of the algorithm before and after improvement are carried out based on the data retrieved from the enhanced magnetic model (EMM). The performance of the improved method is evaluated and verified in the case of the area with normal geomagnetic field (GF), geomagnetic anomaly area, and multiple destinations. The simulation results show that the search efficiency and the straightness of the navigation path are greatly improved. The reason is that the constraint of sample space reduces the randomness in the process of navigation path search, and the improved evaluation function can evaluate the quality of samples more accurately. The improved algorithm also has good performance in the geomagnetic anomaly area, which indicates the potential application in the future.