Volumen 31 (2021): Heft 1 (March 2021)

The paper is devoted to a particular case of the nonlinear and nonautonomous control law design problem based on the application of the optimization approach. Close attention is paid to the controlled plants, which are presented by affine-control mathematical models characterized by integral quadratic functionals. The proposed approach to controller design is based on the optimal damping concept firstly developed by V.I. Zubov in the early 1960s. A modern interpretation of this concept allows us to construct effective numerical procedures of control law synthesis initially oriented to practical implementation. The main contribution is the proposition of a new methodology for selecting the functional to be damped. The central idea is to perform parameterization of a set of admissible items for this functional. As a particular case, a new method of this parameterization has been developed, which can be used for constructing an approximate solution to the classical optimization problem. Applicability and effectiveness of the proposed approach are confirmed by a practical numerical example.

The constrained regulation problem (CRP) for fractional-order nonlinear continuous-time systems is investigated. New existence conditions of a linear feedback control law for a class of fractional-order nonlinear continuous-time systems under constraints are proposed. A computation method for solving the CRP for fractional-order nonlinear systems is also presented. Using the comparison principle and positively invariant set theory, conditions guaranteeing positive invariance of a polyhedron for fractional-order nonlinear systems are established. A linear feedback controller model and the corresponding algorithm of the CRP for fractional nonlinear systems are also proposed by using the obtained conditions. The presented model of the CRP is formulated as a linear programming problem, which can be easily implemented from a computational point of view. Numerical examples illustrate the proposed method.

This paper proposes a fault detection method by extracting nonlinear features for nonstationary and stationary hybrid industrial processes. The method is mainly built on the basis of a sparse auto-encoder and a sparse restricted Boltzmann machine (SAE-SRBM), so as to take advantages of their adaptive extraction and fusion on strong nonlinear symptoms. In the present work, SAEs are employed to reconstruct inputs and accomplish feature extraction by unsupervised mode, and their outputs present a knotty problem of an unknown probability distribution. In order to solve it, SRBMs are naturally used to fuse these unknown probability distribution features by transforming them into energy characteristics. The contribution of this method is the capability of further mining and learning of nonlinear features without considering the nonstationary problem. Also, this paper introduces a method of constructing labeled and unlabeled training samples while maintaining time series features. Unlabeled samples can be adopted to train the part for feature extraction and fusion, while labeled samples can be used to train the classification part. Finally, a simulation on the Tennessee Eastman process is carried out to demonstrate the effectiveness and excellent performance on fault detection for nonstationary and stationary hybrid industrial processes.

The vehicular ad-hoc network (VANET) is subject to various attacks because of its dynamic nature and ephemeral character. In VANET, vehicles communicate with each other for safety awareness. The positioning of an unknown vehicle is one of the critical factors to determine the vehicle’s trustworthiness. Although some positioning techniques have achieved a high accuracy level in VANET, they suffer from dynamic noise in real-world environments. This drawback leads to inaccuracy and unreliability during vehicle positioning. In this paper, an optimal innovation based adaptive estimation Kalman filter (OIAE-KF) is proposed. This algorithm offers an alternative solution for the basic Kalman filter and the innovation based adaptive estimation Kalman filter (IAE-KF). The proposed algorithm makes use of fusion of the global navigation satellite system (GNSS) and the inertial measurement unit (IMU) to improve its performance. The OIAE-KF works based on the innovation sequence and involves three steps such as establishing the innovation sequence, applying the innovation property, checking the optimality of the Kalman filter and, finally, estimating process noise (Q) and measurement noise (R). An optimal swapping method is introduced for optimality check. The efficiency of the proposed OIAE-KF method is proved by comparing the predictions of the existing methods such as the IAE-KF. The results show that the OIAE-KF performs better than the existing techniques. It improves the accuracy and consistency in VANET positioning.

Model predictive control (MPC) algorithms are widely used in practical applications. They are usually formulated as optimization problems. If a model used for prediction is linear (or linearized on-line), then the optimization problem is a standard, i.e., quadratic, one. Otherwise, it is a nonlinear, in general, nonconvex optimization problem. In the latter case, numerical problems may occur during solving this problem, and the time needed to calculate control signals cannot be determined. Therefore, approaches based on linear or linearized models are preferred in practical applications. A novel, fuzzy, numerically efficient MPC algorithm is proposed in the paper. It can offer better performance than the algorithms based on linear models, and very close to that of the algorithms based on nonlinear optimization. Its main advantage is the short time needed to calculate the control value at each sampling instant compared with optimization-based numerical algorithms; it is a combination of analytical and numerical versions of MPC algorithms. The efficiency of the proposed approach is demonstrated using control systems of two nonlinear control plants: the first one is a chemical CSTR reactor with a van de Vusse reaction, and the second one is a pH reactor.

In the frame of stochastic filtering for nonlinear (discrete-time) dynamic systems, the unscented transformation plays a vital role in predicting state information from one time step to another and correcting apriori knowledge of uncertain state estimates by available measured data corrupted by random noise. In contrast to linearization-based techniques, such as the extended Kalman filter, the use of an unscented transformation not only allows an approximation of a nonlinear process or measurement model in terms of a first-order Taylor series expansion at a single operating point, but it also leads to an enhanced quantification of the first two moments of a stochastic probability distribution by a large signal-like sampling of the state space at the so-called sigma points which are chosen in a deterministic manner. In this paper, a novel application of the unscented transformation technique is presented for the stochastic analysis of measurement uncertainty in magnet resonance imaging (MRI). A representative benchmark scenario from the field of velocimetry for engineering applications which is based on measured data gathered at an MRI scanner concludes this contribution.

One of the most popular methods in the diagnosis of breast cancer is fine-needle biopsy without aspiration. Cell nuclei are the most important elements of cancer diagnostics based on cytological images. Therefore, the first step of successful classification of cytological images is effective automatic segmentation of cell nuclei. The aims of our study include (a) development of segmentation methods of cell nuclei based on deep learning techniques, (b) extraction of some morpho-metric, colorimetric and textural features of individual segmented nuclei, (c) based on the extracted features, construction of effective classifiers for detecting malignant or benign cases. The segmentation methods used in this paper are based on (a) fully convolutional neural networks and (b) the marker-controlled watershed algorithm. For the classification task, seven various classification methods are used. Cell nuclei segmentation achieves 90% accuracy for benign and 86% for malignant nuclei according to the F-score. The maximum accuracy of the classification reached 80.2% to 92.4%, depending on the type (malignant or benign) of cell nuclei. The classification of tumors based on cytological images is an extremely challenging task. However, the obtained results are promising, and it is possible to state that automatic diagnostic methods are competitive to manual ones.

We propose a decision support framework (DSF) assisting insulin therapy of diabetic children. Our DSF relies on a medical treatment graph (MTG), which models and graphically represents clinical pathways. Using the MTG, it is possible to plan and adapt medical decisions dependent upon the current health state of a patient and the progress of the treatment. Our MTG fits well with the requirements of clinical practice. The presented work is a cooperative effort of researchers in computer science and medicine. The MTG model has been thoroughly tested and validated using real-world clinical data. The usefulness of the approach has been confirmed by physicians.

At present, most high-accuracy single-person pose estimation methods have high computational complexity and insufficient real-time performance due to the complex structure of the network model. However, a single-person pose estimation method with high real-time performance also needs to improve its accuracy due to the simple structure of the network model. It is currently difficult to achieve both high accuracy and real-time performance in single-person pose estimation. For use in human–machine cooperative operations, this paper proposes a single-person upper limb pose estimation method based on an end-to-end approach for accurate and real-time limb pose estimation. Using the stacked hourglass network model, a single-person upper limb skeleton key point detection model is designed. A deconvolution layer is employed to replace the up-sampling operation of the hourglass module in the original model, solving the problem of rough feature maps. Integral regression is used to calculate the position coordinates of key points of the skeleton, reducing quantization errors and calculations. Experiments show that the developed single-person upper limb skeleton key point detection model achieves high accuracy and that the pose estimation method based on the end-to-end approach provides high accuracy and real-time performance.

Due to a continuous increase in the use of computer networks, it has become important to ensure the quality of data transmission over the network. The key issue in the quality assurance is the translation of parameters describing transmission quality to a certain rating scale. This article presents a technique that allows assessing transmission quality parameters. Thanks to the application of machine learning, it is easy to translate transmission quality parameters, i.e., delay, bandwidth, packet loss ratio and jitter, into a scale understandable by the end user. In this paper we propose six new ensembles of classifiers. Each classification algorithm is combined with preprocessing, cross-validation and genetic optimization. Most ensembles utilize several classification layers in which popular classifiers are used. For the purpose of the machine learning process, we have created a data set consisting of 100 samples described by four features, and the label which describes quality. Our previous research was conducted with respect to single classifiers. The results obtained now, in comparison with the previous ones, are satisfactory—high classification accuracy is reached, along with 94% sensitivity (overall accuracy) with 6/100 incorrect classifications. The suggested solution appears to be reliable and can be successfully applied in practice.

We consider a communication network routing problem wherein a number of users need to efficiently transmit their throughput demand in the form of data packets (incurring less cost and less delay) through one or more links. Using the game theoretic perspective, we propose a dynamic model which ensures unhindered transmission of data even in the case where the capacity of the link is exceeded. The model incorporates a mechanism in which users are appropriately punished (with additional cost) when the total data to be transmitted exceeds the capacity of the link. The model has multiple Nash equilibrium points. To arrive at rational strategies, we introduce the concept of focal points and get what is termed focal Nash equilibrium (FNE) points for the model. We further introduce the concept of preferred focal Nash equilibrium (PFNE) points and find their relation with the Pareto optimal solution for the model.

As the traffic volume from various Internet of things (IoT) networks increases significantly, the need for adapting the quality of service (QoS) mechanisms to the new Internet conditions becomes essential. We propose a QoS mechanism for the IoT gateway based on packet classification and active queue management (AQM). End devices label packets with a special packet field (type of service (ToS) for IPv4 or traffic class (TC) for IPv6) and thus classify them as priority for real-time IoT traffic and non-priority for standard IP traffic. Our AQM mechanism drops only non-priority packets and thus ensures that real-time traffic packets for critical IoT systems are not removed if the priority traffic does not exceed the maximum queue capacity. This AQM mechanism is based on the PI^α controller with non-integer integration order. We use fluid flow approximation and discrete event simulation to determine the influence of the AQM policy on the packet loss probability, queue length and its variability. The impact of the long-range dependent (LRD) traffic is also considered. The obtained results show the properties of the proposed mechanism and the merits of the PI^α controller.