The study analyzes the tasks of air objects navigation support. The authors conducted a syn-thesis of a flight route of an unmanned aerial vehicle with the choice of heading angles of flight at turning points using the direction finding signals received from a stationary ground source. The root-mean-square errors are estimated based on the flying tests. These errors determine a location of emitters during multiple estimation of angular coordinates at points separated in space by direction-finding receivers placed on board of unmanned aerial vehicles “Orlan-10”, “Phoenix”, and “Grant-M”. Methods for correcting autonomous flight with increasing and decreasing accuracy in estimating signal arrival directions are studied, which make it pos-sible to minimize positioning errors for a fixed observation time. The study proposes a scheme of a self-excited oscillator on a field-effect transistor for backup com-mand control with automatic heading angle control. A rule is substantiated for constructing a model of a self-excited oscillator that meets the requirements for a minimum standard deviation calculated using an electrical circuit and a given heading angle with a restriction on the physical feasibility of the layout parameters of a given structure. Recommendations for the implementation of the electronic circuit are formulated and the modes of self-excitation of the oscillator are determined.

The article continues the authors’ research on constructing computer simulators for operators of technological control objects. The concept of hierarchical (two-level) mathematical stimulation of operation staff is proposed, which imitates the dynamic features of a man and implements the logical conclusion of decision-making. This resulted in a quality assessment and training correction in implementing both individual and collective operator training. The apparatus of fuzzy logic and E-nets (extended Petri nets) is used when constructing a logical and dynamic model for training purpose.

The article presents a technique for calculating a maximum amount of a substance in a closed system using a step-by-step graph of chemical reactions. A list of probable one- and two-particle reversible reactions is generated based on the substances that may be a part of the reacting system. The list is narrowed down according to the research objectives and conditions of the reacting system. The obtained list corresponds to the incidence matrix, which is suitable for analyzing the graph via a computer. The search for the maximum amount of the component in question is conducted through the comparison of weights of the graph’s edges at each stage of its intermediate reactions. The graph’s weights are calculated based on the kinetic coefficients of reactions. They also determine the dominant reaction. The amount is calculated under the premise that the dominant reaction is equilibrium. The maximum amount of hydroperoxyl radical HO2 in a hydrogen H and oxygen O reacting system is calculated.

The study describes the development of a service for monitoring plants’ growth in an indoor greenhouse using computer vision models, visual data collection with the esp32-cam card, the OV5640 camera, and the YOLO v4 detection model for extracting individual plants from the images. The plants tracking was performed by the DeepSORT library. The study determined the age of plants according to their type in order to identify their growth rate and notify the user when the parameters achieved. The computer vision methods are implemented through the TensorFlow 2 framework, with 99 % of classification accuracy, and Random Forest coefficient of determination of 0.94 for the regression of a plant’s age.

The article describes a regressive mathematical model of passenger turnover in the air transport of the Russian Federation. The following are used as independent variables: the average flight cost in the economy class of an aircraft per 1,000 km, the average monthly nominal accrued wages of employees in a full range of organizations, the average fare for travel in a second-class carriage of an express unbranded long-distance train per 100 km of track, the working population. Model parameters are identified using two alternative methods of regression analysis: mixed estimation and maximum consistency between the calculated and actual values of the output variable. The constructed versions of the model fully correspond to the content meaning of the independent variables included in their composition and have high accuracy. To solve the problems, the researcher should select one of them according to their features, namely, either the tendency to minimize the disrepancies between the predicted and actual values of passenger traffic in the future or the desire to identify future trends in the indicator's dynamics, possibly, with the greater accuracy.

The study describes a technology for integrating schedule data from a web-extension of the information system supporting the educational process into Google Calendar. Using the high-level programming language C#, the Schedule Transformer software was developed as a console and web-application that implements the function of uploading a week’s schedule or that of a certain date range. A study of the structure of ical-files, which support data import in the most popular calendars (Google Calendar and Yandex Calendar), was conducted during implementation. The application can be used to broaden the functional possibilities of information systems aimed at supporting educational process via the ability to upload data in a user-friendly format.

The global maximum power point tracking of a solar power plant in partial shading demands a global optimization. Standard algorithms for tracking of maximum power point do not provide for a maximum global power of a solar power plant during real time mode due to low convergence. A model of maximizing the generated power of a reconfigurable solar power plant was developed as a modified fuzzy deep neural network based on the modified quantum-behaved particle swarm optimizer. This neural network consists of the following: convolutional units, recurrent neural networks, and fuzzy units. By processing the sensor signals and images of the solar array, the set modified fuzzy deep neural network generates a reference voltage and an electrical interconnection matrix of the parallel-serial solar array, maximizing its power under non-uniform insolation. The neural network demonstrates such advantages as robustness, better efficiency, and tracking speed in comparison
with the model of a reconfigurable solar power plant based on the particle swarm optimization.

The study describes two methods of numeral scaling in a modular number system: one which is based on the interval estimation and the other one which uses iterative algorithm of scaling number X by the coefficient K and includes both the stage of base system expansion and the scaling stage itself. The authors demonstrate the examples and results of algorithms operation provided by the programs developed via Python that simulate algorithms execution on a computer. Estimates of the algorithms complexity were defined in order to compare them and to detect the most appropriate ones.

The article suggests a new approach to building a profile of a relief surface defined by text. The method is based on trigonometric polynomials for approximating basic terrain elements and generalized polynomials for interpolation.