Russian Journal of Cybernetics

Happy 80th Birthday and Best Wishes to Prof. Vladimir Betelin, Member of the RAS. Greetings from the NRC “Kurchatov Institute” – SRISA and the Editorial Board, Russian Journal of Cybernetics

2026-07-03T08:09:55+03:00

80th Anniversary of Vladimir B. Betelin

2026-07-03T08:18:15+03:00

Overview: the Organizing Committee of the Mathematical Ideas of Prof. P.L. Chebyshev, their Applications in Natural Sciences and Artificial Intelligence Technologies International Conference Commemorating Chebyshev’s 205th Anniversary (Obninsk, May 14–16)

2026-07-03T08:27:45+03:00

Reliable Input Data as the Foundation for Trustworthy AI Applications

2026-07-06T13:02:06+03:00

we analyzed the fundamental limitations and risks of artificial intelligence (AI), with particular emphasis on large general-purpose models such as GPT. We showed that the era of extensive AI growth is approaching its limits because the supply of high-quality human-generated data is becoming exhausted. We considered AI-related challenges within the framework of ill-posed inverse problems and demonstrated that computational instability is an inherent mathematical property of such problems. We examined the theoretical foundations of the Transformer architecture and showed that its application to domains without formal syntactic structures, such as medicine and industry, lacks rigorous mathematical justification, raising concerns about the reliability of the resulting outputs. We analyzed examples from secondary and higher education and identified signs of stagnation associated with the delegation of intellectual tasks to generative AI systems. Based on our findings, we concluded that future AI development should shift from universal models toward specialized solutions supported by mathematical justification, verification of domain-specific source data, and mandatory human oversight in applications involving critical infrastructure.

Education: The Contribution of the Government, Society, and the Individual

2026-07-06T13:15:19+03:00

we analyzed the educational ideas presented in the monograph by E. P. Velikhov, V. B. Betelin, and A. G. Kushnirenko and examined Prof. Vladimir Betelin’s contribution to the development of Russian computer science, computer engineering, and the system for training scientific and engineering personnel. Drawing on the monograph, historical sources, and our personal recollections, we studied the evolution of school and engineering education in Russia in the context of the country’s scientific and technological development. We also examined the historical experience of Soviet educational reforms and their role in shaping a mass system for training engineers, scientists, and teachers. Particular attention was paid to the introduction of computer science into the school curriculum in the 1980s and its significance for the development of computer science education in Russia. We examined the relationship between government policy, the scientific community, and educational institutions in developing the human resources required for advances in cybernetics, computer engineering, and high-technology industries. We analyzed the emergence and development of mathematical clubs, specialized physics-and-mathematics schools, academic competitions, and school computer science as key components of the system for preparing future engineers, programmers, and researchers. We also considered V. B. Betelin’s views on the role of education in the country’s scientific and technological development and in training specialists for computer engineering, computer science, and high-technology industries.

The Concept of a Platform for Application-Centric Control System Synthesis

2026-07-06T13:51:51+03:00

we proposed the concept of a low-code/no-code platform for automating the construction of process control systems in the oil and gas industry based on ontological modeling. We carried out a comparative analysis of ontology engineering methodologies (Methodology, On-To-Knowledge, Gruber’s ontology design principles, BORO) and ontology description languages (RDF, RDFS, OWL, KIF, CycL, Gellish) to justify the selection of OWL and the Protégé environment as the implementation tool. We described a five-level hierarchical architecture of the platform, which covers physical devices, controllers, data processing, visualization, and business logic. We proposed a generalized algorithm for platform operation, which ensures the sequential transformation of ontological models through BPMN and DFD diagrams into a project meta-description and final program code. We formulated the basic requirements of the concept, including support for domestically produced lowperformance computing systems.

Impermeable Barrier Detection in an Oil Field Using Interference Testing Data from Two Wells Near a Fault. Part 1. Verification of the Proposed Methodology

2026-07-06T13:56:53+03:00

we studied the possibility of detecting an impermeable barrier in an oil field using two-well interference test data. The wells are located on one side of the barrier. In reservoir engineering, an impermeable barrier denotes a hydraulic no-flow boundary that restricts fluid communication in the reservoir. The relevance of this study stems from the fact that the presence of faults significantly affects oil field development efficiency. We studied why one-well test data are insufficient for impermeable boundary detection, as shown in a series of seven studies published by the authors in 2023–2025. This limitation arises from errors in the estimation of formation parameters determined by non-hydrodynamic methods. We showed that the number of such parameters can be reduced using the proposed approach. We performed two-dimensional numerical hydrodynamic simulations of well interference tests for a homogeneous infinite reservoir and for the same reservoir containing an impermeable boundary. We proposed an algorithm for impermeable boundary detection based on interference response analysis. We evaluated the influence of errors in non-hydrodynamic formation parameter measurements on the validity of boundary detection. We showed that the proposed approach enables reliable detection of an impermeable barrier, even in the presence of parameter estimation errors, in contrast to the one-well case. Further comprehensive investigation is planned to assess the practical applicability of the method.

Data Input Methods in Microkernel Operating Systems: A Case Study of the Baget Real-Time Operating System

2026-07-06T14:02:49+03:00

we developed an input subsystem for the Baget microkernel real-time operating system (RTOS). We studied the adaptation of graphical environments, including the X Window System, to real-time systems and identified the non-deterministic behavior of conventional solutions based on the udev/libinput stack as a major challenge. To address this issue, we developed a hot-plug mechanism for input devices that operates independently of udev, is integrated into a dedicated X server input thread, and provides predictable response times. The proposed mechanism relies on periodic scanning of the /dev file system and interaction with the X server application programming interface (API) through callbacks and synchronization mechanisms. We also developed an input subsystem for user applications that abstracts low-level device interfaces and can be used independently of the X server. In addition, we implemented a touchscreen calibration application based on a five-point calibration procedure and the least-squares method to determine the parameters of an affine transformation. The application automatically generates the corresponding X server configuration. We evaluated all developed components through a series of tests. The results confirm their operability, deterministic behavior, and practical applicability, including integration with the Nuklear graphical user interface framework. The study demonstrates the feasibility of adapting complex software systems for use in specialized deterministic environments.

Exact Solution of the Equations for Inhomogeneous Couette–Poiseuille Shear Flow with Rayleigh Friction

2026-07-06T14:11:12+03:00

we obtained a new exact solution of the modified Navier–Stokes equations with linear Rayleigh friction. The solution describes a three-dimensional steady flow of a viscous incompressible fluid in a plane channel and generalizes the classical Couette and Poiseuille flows. The longitudinal velocity component varies linearly with one transverse coordinate, while the corresponding coefficients vary exponentially with the second transverse coordinate. Using water as an example, we performed a numerical analysis for different values of the Rayleigh friction coefficient. We investigated the effect of the friction parameter on the thickness of the near-wall layer and on the ratio between viscous dissipation and energy dissipation caused by Rayleigh friction. The results show that at low values of the friction coefficient, the flow remains close to the classical solutions. At high values of the friction coefficient, thin boundary layers form, and Rayleigh friction becomes the dominant dissipation mechanism. The proposed solution extends the family of exact solutions in fluid dynamics and can be used to model flows in porous media, filtration processes, and geophysical fluid dynamics problems.

Thermal Analysis of a Local Automation Unit in Marine Integrated Electrical Power Systems Using Domestic Components

2026-07-07T10:14:20+03:00

we studied the development of local automation devices (LADs) for marine integrated electric power systems (IEES) under conditions of technological sovereignty requirements in the Russian shipbuilding industry. We focused on LAD prototypes based on the domestic microcontroller K5500VK018 and analyzed thermal constraints associated with their operation in harsh marine environments. We examined heat dissipation under passive ventilation conditions at ambient air temperatures up to 60 ^∘C and evaluated the thermal state of the devices. We also optimized the enclosure design to prevent overheating of critical components.
We used numerical simulation of conjugate heat transfer and fluid dynamics implemented in SolidWorks Flow Simulation. We performed grid convergence analysis, accounted for thermal radiation using the Discrete Transfer method, and incorporated anisotropic thermal properties of printed circuit board materials. We found that the baseline enclosure design caused the K5500VK018 microcontroller to exceed its maximum allowable operating temperature. We proposed a modified enclosure design with additional ventilation openings.
We showed that the modified design increased the mass airflow rate in the heat-generating zone by 57.5% and reduced the maximum microcontroller temperature to 84.48 ^∘C at an ambient temperature of 60 ^∘C (limit: 85 ^∘C). We demonstrated that the proposed design ensures reliable operation of the device without active cooling systems or heat sinks. The results confirm the feasibility of mass production of domestic LADs for marine IEES and indicate improved technological independence in the context of global supply chain instability.

Impedance Matching of a GaAs Microwave Power Amplifier Die to a Ceramic QFN Package

2026-07-07T10:22:04+03:00

we studied impedance matching between a GaAs microwave power amplifier and a ceramic QFN package. We showed that the initial package configuration does not provide adequate impedance matching due to parasitic effects from the package leads, bond wires, and the central ground pad. We based the analysis on electromagnetic simulations of the die–package structure and S-parameter evaluation. We then modified the package geometry by adjusting the ground pad layout, signal leads, and bond-wire interconnects. We showed that introducing a multistage reactive structure improves impedance matching and reduces the influence of parasitic elements without the need for external matching circuits.

Sound Simulation for Object Collisions in Virtual Environments

2026-07-07T10:41:52+03:00

we proposed methods for sound simulation during collisions of virtual objects in virtual environments. The approach is based on the playback of pre-recorded audio signals triggered by collision events, as well as by sliding and rolling friction between objects. We generated sound samples for each pair of interacting object types using the Audacity audio editor. We implemented sound rendering with the open-source Miniaudio library to support 3D audio in a virtual environment. During simulation, sound volume depends on parameters obtained from collision detection and collision response. We computed sound amplitude during impacts using a harmonic oscillator model. We tested the proposed methods in the VirSim virtual environment software package using interactions between a virtual robot and various objects, as well as object-object interactions. The results show that the approach provides consistent and realistic audio feedback for collision and friction events.

A Neural Network Time Series Model to Forecast Atmospheric Carbon Dioxide Concentrations in Central Siberia

2026-07-07T11:00:44+03:00

the rise in atmospheric carbon dioxide concentrations driven by natural and anthropogenic factors is one of the main causes of climate change, making accurate long-term forecasting of atmospheric carbon dioxide concentrations an important scientific challenge. In this study, we compared the performance of several machine learning methods for forecasting atmospheric carbon dioxide concentrations using time series data collected at the ZOTTO Station in Central Siberia in 2009–2022. We evaluated ensemble methods, including random forest and gradient boosting, together with a recurrent neural network based on the Long Short-Term Memory architecture. The predictor variables included fire intensity, a lag variable representing the number of days since the beginning of the observation period, and meteorological parameters, including air temperature, daily maximum and minimum temperatures, dew point temperature, relative humidity, wind speed, atmospheric pressure, and precipitation. The results showed that the Long Short-Term Memory model achieved higher forecasting accuracy on the validation dataset than the ensemble methods, with a 2.87 mean absolute error, .91 RMS error, and 0.83 coefficient of determination. We found that air temperature, relative humidity, dew point temperature, and wind speed were the most important predictors of atmospheric carbon dioxide concentrations.

Human Activity Recognition in Smart Homes Using WSN and IoT

2026-07-07T11:04:43+03:00

we studied the recent development of smart homes, which is driven by advances in wireless sensor networks (WSN) and the Internet of Things (IoT). These technologies enable realtime applications that expand the range of smart home services and improve resource efficiency. Smart home systems require a structured approach to the integration of WSN and IoT components. Despite the availability of technologies, a methodology for the design and integration of hardware and software components in smart home system is still insufficiently developed. We proposed a threestage methodology for the development of smart home systems based on WSN and IoT. We designed the system architecture, defined evaluation criteria, formulated system requirements, and provided a rationale for the selection of individual system components. The proposed approach provides a structured framework for the systematic integration of heterogeneous devices and services in smart home environments.

Ensemble Deep Reinforcement Learning Approach for Portfolio Management in the Russian Equity Market

2026-07-07T11:17:33+03:00

we studied an ensemble-based deep reinforcement learning approach to portfolio management in the Russian stock market. The study aimed to reproduce a baseline ensemble architecture using Russian equity market data, evaluate its transferability, and identify modifications that improve trading performance over a long investment horizon. We implemented an initial model consisting of three decision-making agents and then extended the analysis by incorporating a broader set of features, macroeconomic indicators, risk-adjusted reward functions, and a mechanism for continuous adaptation of one agent during live trading. We trained and evaluated the models on data from 2015 to 2025 for liquid Russian equities, and we conducted the final performance comparison on the out-of-sample period from 2023 to 2025. The results show that the main driver of performance improvement is continuous off-policy training, where one agent is updated using trading data generated by any active agent in the ensemble. The best-performing configuration achieves a cumulative return of 61.7 percent and outperforms passive benchmark strategies in absolute return. However, the results also reveal a structural limitation: a long-only ensemble without an explicit allocation mechanism to low-risk assets does not provide sufficient capital protection during prolonged bear market conditions combined with high interest rates.

A Stochastic Approach to Estimating Gas Field Resources under Input Data Uncertainty

2026-07-07T11:22:45+03:00

we studied an integrated approach to estimating gas field resources by combining statistical methods with dynamic reservoir simulation. We developed a simulation model that accounts for geological and physical properties of the reservoir, as well as fluid flow characteristics. We proposed an algorithm that enables resource estimation under input data uncertainty. We tested the method on real gas fields and compared it with conventional approaches. The results show that the proposed approach improves estimation reliability and performance.

Comparison of Ensemble Machine Learning Methods for Predicting the End of the Frost Period

2026-07-07T11:26:21+03:00

we studied the performance of ensemble machine learning methods for predicting the end of the frost period. We provided a brief overview of the considered ensemble approaches and investigated how different hyperparameter settings and input data configurations affect model training. We applied several tools, including gradient boosting methods (XGBoost, LightGBM, and CatBoost), random forest (scikit-learn), and logistic regression (scikit-learn), compared the resulting models, and assessed their predictive quality. The results show differences in performance across methods and highlight the impact of hyperparameter tuning and input data selection on prediction accuracy.

Time Series Classification Using Recurrence Plots and Deep Convolutional Neural Networks

2026-07-07T11:33:26+03:00

we addressed the problem of time series classification using deep learning methods, where the raw signal is not processed directly. We first transformed the one-dimensional signal into a two-dimensional representation using recurrence plots, which capture when the system revisits similar states and how frequently these returns occur. This transformation is not always intuitive, but it simplifies subsequent processing, since the resulting representation can be treated as an image and processed using convolutional neural networks. We used a standard convolutional architecture consisting of several convolutional layers, followed by pooling layers for dimensionality reduction and a fully connected layer that produces the final prediction. The model does not rely on manually designed features, as features are learned during training. However, the final performance strongly depends on how well the chosen signal representation preserves the relevant structure of the original time series.

Categorization and Evaluation of Software Used at Surgut State University under the Import Substitution Policy

2026-07-07T11:37:08+03:00

this study addresses the need to ensure the technological independence of Russian universities in the context of sanctions and the implementation of the national import substitution policy. We examined the problem of assessing the functional completeness of software used at Surgut State University. The study aims to develop a software classification framework and a set of assessment criteria that account for the requirements of critical information infrastructure. We analyzed the ISO/IEC 25000 and GOST R ISO/IEC 25010-2015 standards, as well as current approaches to software classification, and developed an assessment methodology that combines quantitative and qualitative metrics. We applied the proposed methodology to the R7-Office software suite and evaluated its functional completeness. The results demonstrate the applicability of the proposed approach to software assessment. The proposed assessment criteria can support decision-making during the transition to domestic software at higher education institutions.

A Method for Building Adaptive Learning Paths for School Students Based on Academic Performance in Standardized School Exams

2026-07-07T11:39:07+03:00

we studied personalized learning in school exam preparation settings and developed a method for constructing adaptive learning paths based on student performance in standardized school exams. The approach targets individual learning needs and uses modern data-driven techniques to support instructional planning.
We collected student performance data and applied Item Response Theory (IRT) to model test responses. We estimated the probability of a correct answer using a three-parameter logistic model. We then inferred student ability levels using maximum likelihood estimation.
We grouped students according to their estimated ability using the K-means clustering method. The resulting clusters provided a basis for adapting learning trajectories to different performance levels.

Optimization of Evacuation Systems in Modern Multifunctional Buildings

2026-07-07T11:42:04+03:00

we proposed a methodology for modeling and optimizing evacuation systems in modern multi-use buildings, taking into account how occupants are distributed across floors and how evacuation routes can be dynamically updated to reflect changing conditions. We reviewed existing approaches to occupant estimation, graph-based and agent-based movement models, and methods for integrating sensor data into dynamic route planning. We performed numerical simulations of several evacuation scenarios. The results show that adaptive routing reduces total evacuation time and helps prevent local crowding. We also discussed practical implementation issues, including system visualization, fault tolerance, and compliance with safety regulations. Finally, we provided recommendations for the design and operation of evacuation systems in modern buildings.