The paper considers the problem of interval observer design for nonlinear dynamic systems described by discrete-time models under external disturbances, measurement noise, and parametric uncertainties. The problem is to design the observer with fewer dimensions than that of the original system; such an observer must generate upper and lower bounds of admissible values of the prescribed nonlinear function of the original system state vector. To solve the problem, special mathematical tool is used. The advantage of this tool is that it allows studying the systems described by models with non-smoo th nonlinearities. To construct interval observer, the reduced-order model of the original system insensitive or having minimal sensitivity to the disturbances is designed. The designing procedure is based on two algorithms: the first one is intended to design the model of minimal sensitivity; the second one is used to reduce the dimension of the model. The rules are formulated to ensure stability based on the prescribed set of the desirable eigenvalues and feedback. The interval observer consists of two subsystems: the first one generates the lower bound, the second one the upper bound. The relations describing both subsystems are given. To construct such an observer in the nonlinear case, the terms of positive and negative influence of variables describing the model are introduced. These terms allow finding out how the upper and lower bounds of these variables will appear in the interval observer. The conditions under which the observer can be designed are given. The theoretical results are illustrated by an example of three tank system. Simulation results based on the package Matlab show the effectiveness of the developed theory.

Study of dynamics of complex networked systems is one of the relevant problems. Networked systems can be in various states, ranging from complete synchronization, when all systems in the network are coherent, to complete desynchronization, i.e. complete incoherence in the functioning of systems. Synchronization phenomenon has already been well studied, namely, the mathematical definitions of synchronization are introduced, algorithms of studying synchronization are proposed, and synchronization conditions of various types of networked systems are established. Whereas a few works are devoted to the study of desynchronization nowadays. This paper introduces output desynchronization notion for networks of nonlinear systems. The definitions about Yakubovich oscillatority are considered and the link between oscillatority and desynchronization in networks of excitable nonlinear systems is established. Excitable systems are stable; therefore, they do not generate oscillations. Adding couplings between such systems can lead to occurrence of oscillations. The conditions about oscillatority in diffusively coupled networks of FitzHugh-Nagumo systems, which are the simplest neuron models, are derived. Firstly, the case of the simplest network of two coupled systems is considered, and afterwards, obtained result is generalized for the case of several systems. Laplace matrix spectrum plays crucial role in dynamics of such networks. The condition that connects the parameters of the uncoupled system in the network and the eigenvalues of the Laplace matrix, is obtained which determines whether the network is oscillatory or not. The number of systems that generate oscillations in such a network depends on the number of eigenvalues of the Laplace matrix that satisfy the obtained conditions. Obtained analytical results are confirmed by simulation. The results of simulation of complete desynchronization in the network, when all systems begin to oscillate, as well as a chimera-like state, in which only a part of the systems oscillates, while the other part are rest, are presented.

In this paper, we consider the problem of determining the hand with which the subject intends to make a movement according to the signals of the electroencephalogram. The relevance of the task is due to the wide spread of brain-computer interfaces, where electroencephalography is one of the main non-invasive methods for obtaining signals from the brain. To solve the problem, temporal and frequency features are selected from the segments of signals preceding the movement, which are fed to the input of the classification machine learning model. In contrast to the standard supervised learning setup, it is assumed that there is no predefined training data set and the training samples for the model are received one after another. Thus, a situation is simulated in which the model must work with a new subject and adjust to them in real time. The traditional method for training linear models in such a paradigm is stochastic gradient descent. Previously, it was shown that the "Stripe" algorithm developed by Yakubovich for a certain problem has a higher convergence rate than stochastic gradient descent. However, this is achieved by performing algorithm step on each feature of the sample. Thus, that version of "Stripe" is not suitable for working with high-dimensional data. This article discusses another version of "Stripe" that does not have this drawback. It is shown that the proposed algorithm has a higher rate of one learning step compared to traditional linear models based on stochastic gradient descent on the BCI competition II dataset.

The concept of "Digital Transformation 2030", which defines the national goals and strategic objectives of the development of the Russian Federation for the period up to 2030, specifies specialized goals and objectives that are an important message for the introduction of intelligent information management technologies in the electric power industry. The main challenges for the transition to digital transformation are the increase in the rate of growth of tariffs for the end consumer, the increasing wear and tear of the network infrastructure, the presence of excessive network construction and the increase in requirements for the quality of energy consumption. The determining factor in the possibility of developing an effective energy policy is the forecasting of electricity consumption using artificial intelligence methods. One of the methods for implementing the above is the development of an artificial neural network (ANN) to obtain an early forecast of the amount of required (consumed) electricity. The obtained predictive values open up the possibility not only to build a competent energy policy by increasing the energy efficiency of an energy company, but also to carry out specialized energy-saving measures in order to optimize the organization’s budget. The solution to this problem is presented in the form of an artificial neural network (ANN) of the second generation. The main advantages of this ANN are its versatility, fast and accurate learning, as well as the absence of the need for a large amount of initial da-ta for a qualitative forecast. The ANN itself is based on the classical neuron and the error back-propagation method with their further modification. The coefficients of learning rate and sensitivity have been added to the error backpropagation method, and the coefficient of response to anomalies in the time series has been introduced into the neuron. This made it possible to significantly improve the learning rate of the artificial neural network and improve the accuracy of predictive results. The results presented by this study can be taken as a guideline for energy companies when making decisions within the framework of energy policy, including when carrying out energy saving measures, which will be especially useful in the current economic realities.

The actual problems of artificial intelligence related to the development of tools for abstract thinking of autonomous intelligent mobile systems are being solved, which allow planning purposeful behavior in hard-to-reach and aggressive environments for humans. Cognitive tools are proposed that provide intelligent systems with the ability to organize purposeful multi-stage activities related to solving complex problems, when a behavior plan is automatically built in some conditions of a problematic environment, and a given behavior goal is achieved in other operating conditions that are beyond the resolution of technical vision. An important feature of the proposed typical elements of knowledge representation and processing is that they allow intelligent systems to organize the output of solving complex problems, relying only on the data stored in the knowledge representation model and coming from the current operating conditions. In the general case, the developed knowledge model of intelligent systems for various purposes consists of declarative and procedural typical elements of their representation. For a formal description of typical elements of declarative knowledge representation, traditional semantic networks and various sets of restrictions are used, reflecting additional conditions for the future functioning of autonomous mobile intelligent systems. As for the formal description of the typical elements of the representation of procedural knowledge, regardless of a specific subject area, fuzzy semantic networks are used for this. This allows autonomous intelligent mobile systems to adapt to specific operating conditions in underdetermined problematic environments and perform complex tasks formulated by them on this basis. The practical significance of the results obtained lies in the effectiveness of their use for the development of intelligent problem solvers that provide autonomous intelligent mobile systems for various purposes with the ability to perform complex tasks in a priori underdetermined problematic environments by adapting the purposeful activity plan formed in general form to specific current operating conditions.

The article studies a trajectory planning task for a group of UGVs with a consideration of wheels-terrain adhesion variation. Within this paper a brief analysis devoted to existing trajectory planning is done. It outcomes with a conclusion of a necessity to produce additional research within this topic. This paper suggests to use a Sampling-based method to solve this trajectory planning task. An algorithm of rapidly exploring random tree (RRT) is used as a basic algorithm. An advantage of this method (typical for Sampling-based methods) is a simplicity of various non-linear restrictions introduction (e.g. obstacles, differential restrictions etc.). In addition we should mention good potential for algorithm parallelization, because of tree structure of the algorithm. However there exists a shortage of the proposed methods — a high consumption of computational resources, and as an outcome a long calculus duration. This paper proposes to overcome this shortage via distributing of computation among UGVs — actors of a group. This is followed by a comparative analysis of distributed and centralized methods. Analysis shows that the main advantage of proposed method is that it can use almost all models of interaction between wheel and terrain. The latter can act a component for calculation of restrictions for motion acceleration over certain types of terrain. Within this paper we did not study models of interaction between wheel and terrain, but instead used empirical data of allowed values of tangential and normal accelerations for specific UGVs in particular conditions. In final part we present results of simulation witch confirm effectiveness of proposed methods.