Siberian State University of Telecommunications and Informatics

We consider a discrete-continuous predator-prey model and a derived discrete model of isolated population. Unlike the well-known Lotka–Volterra model [32], we assume that new individuals are generated at fixed moments. Thus, we obtain a mathematical model represented by a system of ordinary differential equations with impulses. From this system we obtain a model of isolated population in the form of a nonlinear difference second-order equation and study its dynamic regimes and phase changes in both conservative and nonconservative cases. The model is relevant because it agrees with experimental data from freely distributed databases on populations.

This paper deals with a topic within the field of wireless telecommunications-cognitive radio systems. These systems provide best-effort service, enabling unlicensed users to access part of the radio spectrum. Many users, regardless of the technology or radio spectrum frequencies used, face limited resources such as radio spectrum or computing time. Technologies such as cognitive radio provide a specific approach using “spectrum white spots” for secondary (unlicensed) users. The cognitive radio method allows secondary users to avoid interference, capture free channel timeslots, and transmit and receive data. Many algorithms have been proposed for best-effort service, including machine learning algorithms (a subset of artificial intelligence). As a result of our investigations, we propose the same system with a software-defined radio device prototype to analyze experimental data. This paper examines the results of the most effective machine learning with a reinforcement algorithm received from the software-defined radio board. In addition, we consider the peculiarities of implementing cognitive radio on an experimental board and analyze the probability of spectrum occupancy in different design models.

The volume of information transmitted in telecommunication networks is increasing from year to year, and at the same time the use of fiber-optic cable as a guiding medium is growing. In the last twenty to thirty years alone, more than two million kilometers of fiber-optic cable have been laid in the world. Using a fiber-optic cable allows the high-speed data transmission, which is especially important in big data conditions characterized by traffic volumes of 150 gigabits or more per day. It is no longer enough to use a single fiber-optic line to transmit such a volume of information at an acceptable time. The article considers a system of three fiber-optic lines for the Big Data transmission. A special feature of such a sys-tem is the change in bandwidth requirements over time, when only two or even one fiber-optic line is required for traffic transmission at certain time intervals, which allows providing the increased reliability measures during these time intervals. The term "bandwidth on demand" implies providing data transfer speeds depending on customer requests, their applications, and the capabilities of the data transmission system used. The following reliability measures are calculated for the considering system: the failure probability, the total average time spent in up states, the availability factor, the average number of failures over a given time interval, and the average operating time before failure. The mathematical method of Markov processes with incomes is used for calculations.

Modern data centers (DCs) face the problem of high energy consumption, which leadsto increased operating costs and negative environmental impact. This raises the need to developintelligent control systems (ICS) that can improve energy efficiency and optimize the use ofengineering infrastructure.The aim of this work is to evaluate the effectiveness of integrating neural networks and a robustPID controller into an IMS to improve control adaptability, reduce energy consumption andincrease the resilience of data centers to changing operating conditions.Methods: in order to achieve the set objectives, a simulation model of the TIER IV level datacenter was developed in the TRNSYS environment, which allows to simulate dynamicprocesses of energy consumption. The model uses machine learning algorithms to predictthermal load and power consumption, implemented using a neural network trained in MATLABenvironment. A robust PID controller is also implemented to control cooling systems based onthe predicted data. An economic analysis of the efficiency of IMS implementation with thecalculation of key indicators: ROI, NPV and BCR.The novelty of the work consists in proposing an approach to balancing the input data forneural networks, which allows to reduce the spread of amplitudes of oscillations, reduce theprobability of overtraining and increase the generalization ability of the network. For the firsttime, a method of integrating neural networks and robust PID controller for data center energymanagement, taking into account dynamic infrastructure changes, has been developed.Result: The proposed control system reduced the PUE by 3.5%, reduced the powerconsumption of the cooling systems to 40% of the total power consumption of the data centerand improved the accuracy of heat load forecasting. This increased the adaptability of thecontrol and provided a reduction in operating costs.Practical significance: the developed methods and models are applicable for modernization ofdata center engineering infrastructure in order to improve their energy efficiency, reduce costsand ensure environmental sustainability. The results show the feasibility of implementing IMSto improve the resilience and adaptability of data centers to changing operating conditions,which helps to reduce operating costs and environmental impact.

Adam (Adaptive Moment Estimation) is a well-known algorithm for the first-order gradient-based optimization of stochastic objective functions, based on adaptive estimates of lower-order moments. As shown by computational experiments, with an increase in the degree of conditionality of the problem and in the presence of interference, Adam is prone to looping, which is associated with difficulties in step adjusting. In this paper, an algorithm for step adaptation for the Adam method is proposed. The principle of the step adaptation scheme used in the paper is based on reproducing the state in which the descent direction and the new gradient are found during one-dimensional descent. In the case of exact one-dimensional descent, the angle between these directions is right. In case of inexact descent, if the angle between the descent direction and the new gradient is obtuse, then the step is large and should be reduced; if the angle is acute, then the step is small and should be increased. For the experimental analysis of the new algorithm, test functions of a certain degree of conditionality with interference on the gradient and learning problems with mini-batches for calculating the gradient were used. As the computational experiment showed, in stochastic optimization problems, the proposed Adam modification with step adaptation turned out to be significantly more efficient than both the standard Adam algorithm and the other methods with step adaptation that are studied in the work.

In recent twenty years fiber optic sensors using Mandelstam Brillouin scattering effect have obtained a wide application in distributed measurements of temperature and deformation. Their application in engineering structures requires the analysis of functional parameters under conditions of noise interference onto backscatter signal spectrum. The research presents a mathematical model for measurement process based on stimulated scattering emphasizing on the measurement errors analysis. The formulas are presented for evaluation of mean quadratic errors in temperature and deformation measurements depending on noise-to-signal ratio and number of measurement cycles. In addition, methods to assess spatial resolution and dynamic range are proposed. Variations in methods for detection and processing backscattering spectra make it possible to produce efficient sensor systems which can monitor parameter change in real time.

The task before the developers was to create a geolocation system using Russian manufacturer equipment, so that the error in location determination would not exceed 5 meters. To implement a local positioning system, the pilot area for conducting experiments was defined, a Wi-Fi test network was designed and built in the course of the work on the project. The signal strength level at the input of the receiving device was chosen as a physical parameter, the measurement of which is necessary to calculate the target object’s coordinates. A combined location determitation algorithm based on the mechanisms of nearest access point, spatial pattern differentiation and trilateration was developed. More than a thousand measurements were carried out in the test network, and during that the main problems of the location system were identified and ways to overcome them were proposed

Relevance. The use of multi-antenna technologies in the form of signal precoding is a basic condition for increasing spectral efficiency in modern mobile communication systems in combination with the transition to the millimeter wave frequency band. Propagation conditions in the millimeter wave require the use of antenna arrays in communication systems to compensate for propagation losses and directional transmission and reception of user signals. When transmitting multiple parallel spatial user data streams, signal precoding is used to implement spatial multiplexing and improve the spectral efficiency of the system. A hybrid architecture for constructing a multi-antenna system and precoding, consisting of analog and digital parts, is considered. But reducing the number of radio frequency paths leads to a decrease in the possibility of spatial multiplexing compared to a completely digital system. In this regard, it is important to select the optimal number of radio frequency paths to obtain maximum spatial multiplexing, taking into account the current conditions of signal propagation and spatial correlation of the communication channel. The purpose of the study is to determine the effect on the spectral efficiency of the choice of the number of used radio frequency paths in a hybrid precoding system.The research methods consist of simulation modeling of a hybrid precoding algorithm. To solve the problem of numerical modeling of hybrid precoding, implementations of a MIMO channel of millimeter waves are used, obtained using the open channel model software package QuaDRiGa.The results are presented in the form of distribution functions of the spectral efficiency of the hybrid precoding system, obtained on the basis of channel implementations in a certain propagation scenario.The novelty lies in the numerical determination of the channel parameters of a multi-antenna communication system in the millimeter range and the use of the distribution of eigenvalues of the obtained channel implementations to study the influence of the number of radio frequency paths on the spectral efficiency of a communication system with hybrid precoding achievable under given conditions.

In the modern circumstances related to the public life's informatization and digitalization, as well as the development of infocommunication technologies, the greatest interest lies in the services based on determining the object's location. The developers are facing the topical challenge of introducing the artificial intelligence to the geolocation technologies. Local positioning systems allow navigating or searching for objects indoors, where the radio signal from satellite (global) navigation systems reaches being greatly distorted and does not provide acceptable geolocation accuracy. The authors have developed the software intended for determination of the mobile object's coordinates with the use of neural networks. Two different approaches to this issue are compared in this article - firstly, the use of Neural Network Toolbox from the MATLAB application package, and secondly, the development of software modules in the Python programming language using the TensorFlow and Keras libraries. The authors analyzed the obtained results by such criteria as, first of all, the highest accuracy (least error) of positioning, as well as ease of implementation and adaptability of the proposed solutions.

The results of the numerical investigation of the hybrid precoding performance are presented under typical deploynment scenario of single user multi-stream wireless communication. Millimeter-wave systems employ the antenna arrays to compensate the high channel path loss by the antenna array gain. Multi stream data precoding can be used to implement spatial multiplexing and enhance the system spectral efficiency. The fully digital precoding in mmWave band has the high hardware cost and power consumption. The hybrid analog and digital precoding allow to use the reduced number of RF chains in combination with baseband digital precoding. The hybrid precoding approximates the optimal digital precoding matrix by product of a baseband digital precoding matrix and an analog precoding matrix implemented by phase shifters. The hybrid approximation of optimal precoding matrix is found as matched to the sparse spatial properties of multipath mmWave wireless channel. For numerical simulation of the performances of hybrid precoding the QuaDRiGa channel model is used.