Data Dropouts Research Articles

Automatic Recognition of Soybean Leaf Diseases Using UAV Images and Deep Convolutional Neural Networks

Plant diseases are a crucial issue in agriculture. An accurate and automatic identification of leaf diseases could help to develop an early response to reduce economic losses. Recent research in plant diseases has adopted deep neural networks. However, such research has used the models as a black-box passing the labeled images through the networks. This letter presents an analysis of the network weights for the automatic recognition of soybean leaf diseases applied to images taken straight from a small and cheap unmanned aerial vehicle (UAV). To achieve high accuracy, we evaluated four deep neural network models trained with different parameters for fine-tuning (FT) and transfer learning. Data augmentation and dropout were used during the network training to avoid overfitting. Our methodology consists of using the SLIC method to segment the plant leaves in the top-view images obtained during the flight. We tested our data set created from real flight inspections in an end-to-end computer vision approach. Results strongly suggest that the FT of parameters substantially improves the identification accuracy.

HANKEL‐EM‐SVD: a hybrid data dropout estimation technique for high voltage partial discharge signals

Partial discharge (PD) measurement and characterisation is the most effective method to assess the insulation conditions of equipment for its diagnosis. A significant challenge in this area is PD denoising. Despite recent advancements in denoising algorithms, data drops out because of inappropriate threshold selection during denoising. Processing PD signals with missing data leads to an inaccurate assessment of the insulation conditions, indicating recovering such missed data is a potential research scope under the moonlight. This study addresses the issues of recovering such data dropouts by considering the PD data vector matrix as a low-rank matrix that needs completion by Hankel-matrix decomposition methods. After finding the probabilistic estimate of the missing values using an expectation maximisation algorithm, singular-value decomposition is used to find the actual missing values by soft-thresholding the singular values of the matrix. After testing this technique on simulated PD data, it is implemented to test SASTRA-High-Voltage Laboratory data, showing root mean square error (RMSE) 0.77% and mean absolute error 0.84%. The efficiency of this technique is confirmed when tested with large-sized noise free PD data of 36 samples from the Laboratory of BOLOGNA University (with RMSE varying from 0.15 to 0.31% and mean absolute error varying from 0.29 to 0.6%).

Robust Combined Lane Keeping and Direct Yaw Moment Control for Intelligent Electric Vehicles with Time Delay

In order to enhance the tracking performance and improve the stability of intelligent electric vehicles, combined Lane keeping and direct yaw moment control is a good choice. In this paper, an uncertain model of intelligent electric vehicles for combing the lane keeping and direct yaw moment control is deduced, in which time delay and data dropouts are involved. Since the intelligent electric vehicles have the features of time delay and strong uncertainties, a novel robust guaranteed cost combined lane keeping and direct yaw moment control system is constructed to manage the lateral motion of intelligent electric vehicles. The asymptotic stability of combined lane keeping and direct moment control system is verified based on the Lyapunov stability theory. Simulation tests are carried out to demonstrate the feasibility of the proposed control approach, and the results indicate that the presented robust combined control approach can accurately achieve the lane tracking capability, stability and maneuverability of intelligent electric vehicles.

Networked and Cloud Control Systems - Modern Challenges in Control Engineering

The paper describes development of some control techniques used in the feedback control systems. Modern control trends,with attention to networked control systems (NCS), and basic assumptions for introducing a new control concept – control in the cloudare described. Networked control systems techniques initiated the development of concept of Internet of Things and are fundamental incloud control systems design. Network induced delays that occur in NCSs, as a result of the presence of a communication network, andNCSs stability are considered, as well. Time delays and data dropouts influence on the NCS behavior and stability is analyzed, and somerelations for maximum allowable delay bound estimation are provided. It is shown, on example of DC motor control, that networkedsystem behaves in desired manner and remains stable if the delay is less then estimated maximum bound. When data dropouts occur,system delays become greater and lead to the system instability. Cloud control systems use some techniques of NCSs and cloudcomputing. These techniques are briefly presented and are necessary to complete control tasks in the cloud.

Palm vein recognition based on end-to-end convolutional neural network

We propose a novel palm-vein recognition model based on the end-to-end convolutional neural network. In this model, the convolutional layer and the pooling layer were alternately connected to extract the image features, and the categorical attribute was estimated simultaneously via the neural network classifier. The classification error was minimized via the mini-batch stochastic gradient descent algorithm with momentum to optimize the feature descriptor along with the direction of the gradient descent. Four strategies including data augmentation, batch normalization, dropout, and L2 parameter regularization were applied in the model to reduce the generalization error. The experimental results showed that for classifying 500 subjects form PolyU database and a self-established database, this model achieved identification rates of 99.90% and 98.05%, respectively, with an identification time for a single sample less than 9 ms. The proposed approach, as compared with the traditional method, could improve the accuracy of palm vein recognition in clincal applications and provides a new approach to palm vein recognition.

A comparison of compensation methods for random input data dropouts in networked iterative learning control system

In this paper, for a class of linear networked iterative learning control (ILC) systems, methods to compensate dropped input data in time or iteration domain are compared. Specifically, the transition matrices of input error at the controller side with the two methods are derived first, respectively. After that, the varieties of eigenvalues and elements in the lower triangular of the transition matrices are analyzed. Through analyzing the varieties, it can be easily found that the two methods guarantee the convergence of input error at the controller side, while only the compensation in iteration domain guarantees the convergence of input error at the actuator side. Due to the introduction of networks, the convergence of output error is determined by the input error at the actuator side. Hence, a conclusion could be made naturally that the output error converges to zero with compensation in iteration domain, while compensation in time domain cannot guarantee that. Finally, numerical experiments are given to corroborate the theoretical analysis.

Glaucoma Diagnosis with Machine Learning Based on Optical Coherence Tomography and Color Fundus Images.

This study aimed to develop a machine learning-based algorithm for glaucoma diagnosis in patients with open-angle glaucoma, based on three-dimensional optical coherence tomography (OCT) data and color fundus images. In this study, 208 glaucomatous and 149 healthy eyes were enrolled, and color fundus images and volumetric OCT data from the optic disc and macular area of these eyes were captured with a spectral-domain OCT (3D OCT-2000, Topcon). Thickness and deviation maps were created with a segmentation algorithm. Transfer learning of convolutional neural network (CNN) was used with the following types of input images: (1) fundus image of optic disc in grayscale format, (2) disc retinal nerve fiber layer (RNFL) thickness map, (3) macular ganglion cell complex (GCC) thickness map, (4) disc RNFL deviation map, and (5) macular GCC deviation map. Data augmentation and dropout were performed to train the CNN. For combining the results from each CNN model, a random forest (RF) was trained to classify the disc fundus images of healthy and glaucomatous eyes using feature vector representation of each input image, removing the second fully connected layer. The area under receiver operating characteristic curve (AUC) of a 10-fold cross validation (CV) was used to evaluate the models. The 10-fold CV AUCs of the CNNs were 0.940 for color fundus images, 0.942 for RNFL thickness maps, 0.944 for macular GCC thickness maps, 0.949 for disc RNFL deviation maps, and 0.952 for macular GCC deviation maps. The RF combining the five separate CNN models improved the 10-fold CV AUC to 0.963. Therefore, the machine learning system described here can accurately differentiate between healthy and glaucomatous subjects based on their extracted images from OCT data and color fundus images. This system should help to improve the diagnostic accuracy in glaucoma.

Iterative consensus control for a class of nonlinear MIMO multi-agent systems with data dropout

Consider frequent data dropout phenomena of measurements due to communication channel failure. The consensus tracking problem for a class of nonlinear MIMO multi-agent systems (MAS) with data dropouts is addressed in this paper. The phenomena of data dropout is described by an introduced random variable, which is defined with a Bernoulli sequence valued 0/1. Moreover, the situation which only part of the agents can access the desired trajectory information also take into account. Based on the random variable, a modified distributed P-type iterative learning control algorithm with data drops is proposed. In addition, the identical initialization condition of classic iterative learning control (ILC) is released as well. Then, the theoretical analysis of the system convergence condition is given by using the contraction mapping method and supermom norm technical. As a result, the proposed ILC algorithm can guaranteed the system convergence with data dropout in a finite time. Finally, an illustrative example is presented to demonstrate the perfect tracking performance and the effectiveness of the scheme.

Input-to-State Stabilization of Nonlinear Systems via Event-Triggered Impulsive Control

In this work, an event-triggered impulsive control (ETIC) method is proposed to address the input-to-state stabilization (ISS) issue for a class of nonlinear systems under external disturbance. Different from many existing impulsive controllers usually with fixed impulse instants, we design a novel one which is event-triggered and multimodule constrained. Based on that, we establish some ISS criteria for the considered system with and without data dropouts. The presented event-triggered control method can save the communication resources and avoid the Zeno behaviors. Finally, an example with simulations results is provided to illustrate the validity of the method.

Wide-Area Damping Control using Reduced Copy under Intermittent Observation: A Novel Performance Measure

Insight into the application of observer-driven reduced copy (ORC) to deal with data packet dropout during wide-area oscillation damping control through wind farms (WFs) is developed. The ORC technique tries to emulate the behavior of the power system by using the information of the nominal dynamics under intermittent observation. This brief presents a direct performance measure for ORC. To that end, a standard lemma relating the bound on the norm of the integral over the product of two functions is presented, which is proven using Holder’s inequality. Based on the lemma, an expression is derived for the bound on the norm within an inter-sample interval for difference between the power system states in presence of data dropout and that under ideal communication. Such a bound, when evaluated in a relative sense, would be useful for system operators to perform online contingency screening and ranking from damping performance standpoint without running prohibitively expensive time-domain simulations. In addition, such a bound is also derived for the conventional feedback control (CFC). The data dropout is modeled using the Gilbert-Elliott model. The analytical results are validated using time-domain simulation of a 16-machine, 5-area New England-New York system with two doubly fed induction generator-based WFs.

Distributed Conflict-Detection and Resolution Algorithm for UAV Swarms Based on Consensus Algorithm and Strategy Coordination

In this paper, we study the problem of conflict detection and resolution for unmanned aerial vehicle (UAV) swarms. Specially, we propose a distributed conflict detection and resolution method for multi-UAVs in formation based on consensus algorithm and strategy coordination. When encountering threat swarms, the UAVs in one swarm act as one unit and are together treated as one control object. Each swarm in conflict selects three candidate collision avoidance maneuvers from the preset strategy pool, generates the corresponding planned trajectories with an uncertainty trajectory modeling, and then broadcasts and shares them. All of the swarms in conflict coordinate and determine an optimal combination of strategies. When a collision is imminent, the primary strategy is activated. Each swarm adopts a “leader-follower” strategy, where the leader UAV is regarded as the controller and flies independently, and the others follow the leader UAV. During motion, a decentralized consensus algorithm is adopted for agents to converge to their positions for the desired formation and to maintain a stable geometric configuration. A temporal and spatially integrated conflict-detection model is improved, especially for UAV swarms. A token-allocation strategy is especially improved for distributed coordination to resolve the partial knowledge of the airspace condition. Damping of the coordination is proposed to address data dropouts and transmission delays. Two typical scenarios are conducted to test the methodology proposed in this paper. The simulation result demonstrates the effectiveness and rationality of the proposed methodology.

Distributed Conflict-Detection and Resolution Algorithms for Multiple UAVs Based on Key-Node Selection and Strategy Coordination

A distributed conflict-detection (CD) and resolution algorithm for multiple unmanned aerial vehicles based on key-node selection and strategy coordination are proposed in this paper. Each aircraft is regarded as a critical node in a complex network. A novel temporal and spatial integrated CD methodology is proposed for aircraft formations and clusters. Each node selects three candidate strategies from a pre-set pool of strategies and generates corresponding planned tracks using uncertainty modeling based on the 6 degrees-of-freedom motion. The primary combination of strategies required for coordination of the aircraft in conflict is based on the maximum-robustness principle. To address partial knowledge of the environment, a special token-allocation strategy for coordination with incomplete information is proposed in this paper. To address potential data dropouts, this study damps the solution to achieve coordination. Two extreme scenarios are constructed to examine the proposed methodology. The results showed great collision-avoidance ability in continuous conflicts with multiple aircraft in a highly dynamic 3-D environment.

Networked Fault Detection for Markov Jump Nonlinear Systems

This paper deals with the problem of dissipativity-based asynchronous fault detection (FD) for Takagi–Sugeno fuzzy Markov jump systems with network data dropouts. It is assumed that data dropouts happen intermittently from the plant to the FD filter, which is described by Bernoulli process. The hidden Markov model is employed to describe the asynchronous phenomenon between the plant and filter. Based on Lyapunov theory, a sufficient condition is developed to guarantee that the FD system is stochastically stable with strictly dissipative performance. By choosing an appropriate Lyapunov function with the slack matrix technique and Finsler's Lemma, two approaches are proposed to compute filter gains by solving linear matrix inequalities. Finally, an example is provided to illustrate the usefulness and effectiveness of the proposed design methods.

A Tool to Characterize Delays and Packet Losses in Power Systems With Synchrophasor Data

This paper describes the implementation of a tool to estimate latencies and data dropouts in communication networks transferring synchrophasor data defined by the C37.118 standard. The tool assigns a time tag to synchrophasor packets at the time it receives them according to a global positioning system clock and with this information is able to determine the time those packets took to reach the tool. The tool is able to connect simultaneously to multiple phasor measurement units (PMUs) sending packets at different reporting rates with different transport protocols such as user datagram protocol or transmission control protocol. The tool is capable of redistributing every packet it receives to a different device while recording the exact time this information is re-sent into the network. The results of measuring delays from a PMU using this tool are presented and compared with those of a conventional network analyzer. The results show that the tool presented in this paper measures delays more accurately and precisely than the conventional network analyzer.

Centralized Dynamic State Estimation Using a Federation of Extended Kalman Filters With Intermittent PMU Data From Generator Terminals

An improved dynamic state estimation scheme that performs estimation for the full plant (states of a generator, exciter field voltage, and governor mechanical torque) using intermittent data from a phasor measurement unit (PMU) connected at generator terminal is presented. Overall fourth-order generator model is assumed in an extended Kalman filter (EKF), while first-order governors and excitation systems are assumed for simplicity of large-scale implementation. State estimation is performed using the EKF with random PMU data dropouts and known inputs, i.e., secondary reference signals P ref and V ref provided to a power plant by the network control center from economic dispatch. The state estimation scheme has been extended to all generators in network and DSE is performed using a computationally decentralized federation of EKFs at a centralized phasor data concentrator where PMU data are aggregated while dealing with a specified stochastic dropout rate. Required modifications have, thus, been made to standard EKF formulation to account for communication channel interruption and inherent delays. Simulation studies performed on the benchmark IEEE 9 and 39 bus system demonstrated performance and resilience of the proposed centralized EKF-based estimation technique. We also found that a centralized estimator can lead to improved wide-area instability indices derived from state estimates rather than PMU data directly.

Robust Quantized H∞ ILC Design for Uncertain Systems with Communication Constraints

This paper considers the problem of H∞ iterative learning control design for network-based uncertain systems with communication constraints, including data quantization and random data dropouts. It is assumed that the control input update signals are quantized before they are transmitted to the iterative learning controller and a logarithmic quantizer is used to decode the data with a number of quantization levels. Then, the data dropout is modeled by the conventional Bernoulli random variable to describe the successful transmission or miss. The 2-D dynamic of such ILC process is established by a stochastic Roesser model. By using sector bound method to deal with the quantization error, a sufficient asymptotically stability condition with an H∞ disturbance attenuation level for the 2-D system is given and then the ILC design can be developed based on the condition. The effectiveness of the proposed method is illustrated by application to an injection molding process. DOI: http://dx.doi.org/10.5755/j01.itc.47.3.18451

Iterative learning fault-tolerant control for networked batch processes with event-triggered transmission strategy and data dropouts

ABSTRACTThis paper studies the iterative learning fault-tolerant control (ILFTC) problem for networked batch processes with event-triggered transmission strategy and data dropouts. During the transmission of input signal, the event-triggered mechanism is adopted to reduce the number of updated data items. The data dropouts are assumed to obey the Bernoulli random binary distribution. The objective of this paper is to design a state feedback controller such that the system is fault-tolerant and satisfies the robust performance requirement. By combining 2D stochastic system theory and linear matrix inequality (LMI) technique, some sufficient conditions are given to ensure the existence of the designed controller. Finally, an example of nozzle pressure control is utilized to verify the availability of the proposed method.

Iterative learning control with incomplete information: a survey

This paper conducts a survey on iterative learning control &#x0028 ILC &#x0029 with incomplete information and associated control system design, which is a frontier of the ILC field. The incomplete information, including passive and active types, can cause data loss or fragment due to various factors. Passive incomplete information refers to incomplete data and information caused by practical system limitations during data collection, storage, transmission, and processing, such as data dropouts, delays, disordering, and limited transmission bandwidth. Active incomplete information refers to incomplete data and information caused by man-made reduction of data quantity and quality on the premise that the given objective is satisfied, such as sampling and quantization. This survey emphasizes two aspects: the first one is how to guarantee good learning performance and tracking performance with passive incomplete data, and the second is how to balance the control performance index and data demand by active means. The promising research directions along this topic are also addressed, where data robustness is highly emphasized. This survey is expected to improve understanding of the restrictive relationship and trade-off between incomplete data and tracking performance, quantitatively, and promote further developments of ILC theory.

Improving the performance of networked control systems with time delay and data dropouts based on fuzzy model predictive control

This paper proposes a fuzzy model predictive control (FMPC) combined with the modified Smith predictor for networked control systems (NCSs). The network delays and data dropouts are problems, which greatly reduce the controller performance. For the proposed controller, the model of the controlled system is identified on-line using the Takagi – Sugeno (T-S) fuzzy models based on the Lyapunov function. There are two internal loops in the proposed structure. The first is the loop around the FMPC, which predicts the future outputs. The other is the loop around the plant to give the error between the system model and the actual plant. The proposed controller is designed for controlling a DC servo system through a wireless network to improve the system response. The practical results based on MATLAB/SIMULINK are established. The practical results are indicated that the proposed controller is able to respond the networked time delay and data dropouts compared to other controllers.

Wide-Area Damping Control Using Multiple DFIG-Based Wind Farms Under Stochastic Data Packet Dropouts

Data dropouts in communication network can have a significant impact on wide-area oscillation damping control of a smart power grid with large-scale deployment of distributed and networked phasor measurement units and wind energy resources. Remote feedback signals sent through communication channels encounter data dropout, which is represented by the Gilbert-Elliott model. An observer-driven reduced copy (ORC) approach is presented, which uses the knowledge of the nominal system dynamics during data dropouts to improve the damping performance where conventional feedback would suffer. An expression for the expectation of the bound on the error norm between the actual and the estimated states relating uncertainties in the cyber system due to data dropout and physical system due to change in operating conditions is also derived. The key contribution comes from the analytical derivation of the impact of coupling between the cyber and the physical layer on ORC performance. Monte Carlo simulation is performed to calculate the dispersion of the error bound. Nonlinear time-domain simulations demonstrate that the ORC produces significantly better performance compared to conventional feedback under higher data drop situations.