The subject of this study is various types of network traffic in modern computer networks with a complex structure and a certain degree of self-similarity. Efficient use of network resources and ensuring the quality of service to subscribers are important tasks of computer networks. The probability of losing a message due to buffer storage device overflow is an important parameter in determining the quality of service (QoS). The mathematical model should be used to estimate this parameter. Recent advancements have resulted in many different models of packet loss probability in a router buffer. However, many models do not consider the traffic characteristics of various modern applications and protocols. The traffic in modern computer networks has a complex structure and often has a certain degree of self-similarity. Currently, a large number of models are available for estimating the probability of packet loss due to buffer overflow. The goal of this work is to perform a comparative analysis of such models and provide recommendations for their use and to estimate the influence of network traffic fractality on the probability of packet loss in a router due to buffer overflow. The tasks to be solved are as follows: 1) to conduct an analysis of analytical models that describe the packet loss probability in a router considering the influence of fractality and without it; 2) to construct the dependencies of the packet loss probability in the router on the data transmission channel load for different buffer capacity values, the Hurst exponent, and traffic deviation; and 3) to describe the dependences of the packet loss probability on the buffer capacity for different channel load values. Comparative analysis of various methods of fractal traffic modeling and simulation with different storage capacity, Hurst exponent, deviation coefficients, and channel load factor values is used. The following results were obtained: 1) The M/M/1 queuing system model gives the most optimistic estimate. This estimate can be used as a lower bound for the message loss probability for a given buffer capacity and a channel load factor; 2) the highest message loss probability was observed when using queuing systems with a Hurst exponent of 0.95; 3) the packet loss probability also increased with an increasing traffic fractality and deviation coefficient; 4) the influence of fractality decreased with an increase in the buffer capacity was found; 5) an objective estimation of the message loss probability due to a router buffer overflow can only be made by considering the nature of the traffic. Conclusions. The main contribution of this research is that various types of network traffic have a fractal nature, and the traditional methods of route service specification, such as traffic using the M/M/1 queuing model, give more errors. Because of the research conducted to reduce the impact of traffic fractality, increasing the capacity of buffer storage devices is necessary.
The subject are mathematical models of decision-making under uncertainty in a production environment with human presence. The research objectives: is to form a safe and effective policy for controlling the motion of a three-link collaborative robot-manipulator, by developing a mathematical model of a partially observable Markov decision process (POMDP). Methods: methodology of partially observable Markov processes (POMDP), numerical modeling, approximation of the expected reward, comparative analysis of scenarios with different risk parameters. Results: the implemented model is able to form an adaptive policy for the manipulator's behavior taking into account incomplete information about the person's position; the dependence of optimal actions on the probability distribution of the human position and the intensity of the risk penalty is demonstrated; the influence of the difference in rewards between the fast movement and stop modes on the choice of actions is shown. Conclusions: the developed POMDP model can be used as a basis for building high-level adaptive control of a collaborative manipulator in a shared workspace with a human. The proposed approach has the prospect of being implemented in flexible production systems and cyber-physical complexes, in particular, taking into account dynamic risk reassessment and integration with computer vision algorithms.
The subject matter of the article is the efficiency analysis of greedy optimization algorithms for subset selection in distributed systems under delta-matroid constraints. The goal is to compare the performance of the classical unconstrained greedy algorithm and the GREEDI algorithm with delta-matroid constraints in terms of solution quality, computational characteristics, and scalability. The tasks to be solved are: to implement both algorithms; to perform simulations on synthetic graph datasets with sizes ranging from 10 to 100 nodes; to benchmark computational efficiency and approximation quality; to analyze the impact of delta-matroid constraints on benefit maximization and distributed execution. The methods used are: graph-based modeling, combinatorial optimization under matroid-type constraints, approximation algorithms, and distributed processing frameworks. The following results were obtained: GREEDI consistently provided higher-benefit subsets compared to the unconstrained greedy algorithm, achieving better trade-offs between execution time and solution quality; the distributed processing framework demonstrated scalability for large datasets and supported real-time responsiveness; performance advantages were more pronounced for larger graphs and higher constraint densities. Conclusions. The scientific novelty of the results obtained is as follows: 1) an experimental validation of the GREEDI algorithm under delta-matroid constraints for distributed subset selection was carried out; 2) the influence of such constraints on approximation quality and computational characteristics was quantified; 3) a scalable real-time processing approach for large graph-structured data was proposed, enabling potential applications in sensor deployment, recommendation systems, feature selection, and cache optimization.
The subject of this study is the digital transformation of the occupational safety management system in civil aviation. Owing to the country’s unique geopolitical position in the centre of Eurasia, the Republic of Kazakhstan’s rapid growth in cargo and passenger traffic is associated with increasing employee risks, making it critically necessary to review existing occupational safety management systems. Traditional approaches to occupational safety management, which often focus on reactive elimination of consequences, are insufficiently effective in the context of the dynamic development of a high-tech industry, where every incident has serious economic, reputational, and social repercussions. It is possible to improve occupational safety by applying advanced digital technologies, in particular digital twin and AI agent technologies, which are capable of continuously accumulating, processing, and transmitting vast amounts of data in real time through self-learning, creating a comprehensive picture of an aviation enterprise’s entire occupational safety management system’s functioning. This study aims to justify the feasibility of modifying the occupational safety management system in civil aviation in the Republic of Kazakhstan by integrating digital twin and AI agent technologies into key functional processes. Results. The proposed algorithm for the developed AI agent, which is explicitly designed for integration into the occupational safety management systems of aviation enterprises in Kazakhstan, is the key contribution of this study. Its architecture, operating principles, and algorithms for interacting with big data received from digital twins of various elements of the aviation system are described in detail, ranging from the condition of aircraft and ground equipment to the personnel’s psychophysiological indicators and the working environment’s characteristics. This algorithm enables the AI agent to detect anomalies and build predictive models, signalling potential threats in advance. The results of the AI agent’s risk assessment in the civil aviation occupational safety system have been visualised, demonstrating its high efficiency in identifying vulnerabilities, predicting critical situations and forming informed, personalised recommendations for their prevention. The research results demonstrate how proactive monitoring and analysis performed by an AI agent based on digital twin data can significantly reduce the likelihood of injuries and occupational diseases. Conclusions. The proposed approach to modifying the occupational safety management system at civil aviation enterprises is based on the synergistic integration of digital twins and AI agents, whereby risk management shifts from reactive elimination to preventive modelling and mitigation of potential threats. The creation of an occupational safety management system at the country’s aviation enterprises, based on the use of digital twins and AI agents, will significantly increase the competitiveness of civil aviation in the Republic of Kazakhstan on the world market, positioning it as a leader in the application of high-tech solutions for ensuring occupational safety and sustainable development.
The object of the study is a formalized model of requirements traceability in project documentation for hardware-software systems. The subject matter of the research encompasses the application of mathematical modeling and tool-based approaches to automate the traceability process, focusing on the design and functionality of the TraceTrend software tool. The primary goal of the study is to improve the quality and integrity of requirements management by implementing traceability mechanisms that ensure logical consistency, hierarchical correctness, and complete test coverage across all project documentation stages. The research tasks include: identifying challenges related to manual requirements tracing in safety-critical domains; constructing a formal mathematical model based on set theory, binary relations, and directed graphs; defining binary matrices for requirement inheritance and test coverage; developing automated analysis techniques for traceability conditions; integrating the model into the TraceTrend tool; and demonstrating its applicability through a real-world case study. The study employed the following methods: mathematical modeling of binary relations, model-based testing, static analysis of documentation structures, and the use of Boolean matrix operations for verifying coverage and consistency. As a result of the research, a formal model of requirements traceability was created and implemented in the TraceTrend tool. The tool enables automated extraction of requirement identifiers, construction of traceability matrices, and verification of coverage and logical completeness. The application of TraceTrend has shown its effectiveness in identifying undocumented requirements, broken dependencies, and gaps in test coverage early in the project lifecycle. Conclusions. The integration of formal models and traceability tools significantly strengthens the reliability and auditability of requirements management processes in engineering projects. TraceTrend has proven to be a valuable instrument for improving documentation quality and supporting compliance with standards such as IEC 61508 and ISO/IEC/IEEE 29148. Although the tool requires initial configuration for requirement markup, its benefits in enhancing visibility, consistency, and verification readiness justify its adoption in high-assurance development environments. The study confirms the necessity of embedding formal traceability analysis into standard project workflows to ensure both structural rigor and regulatory compliance.
This integrated approach aims to prevent suboptimal prioritization, ensure effective resource allocation, and expedite the recovery of information systems. Methodology: the proposed methodology establishes a hierarchical, multi-factor prioritization approach. It integrates the quantitative technical severity assessment provided by CVSS with Ukraine’s national criticality levels. This dual-layer scoring is further supplemented by a structured tie-breaking mechanism using additional attributes to achieve precise prioritization. A structured dataset was constructed, covering nationwide impact, economic consequences, information-related impacts, functional consequences, recovery capabilities, and system/network specifics. A prioritization methodology was developed, involving five key phases: incident registration, data verification, correlation and aggregation, criticality assessment, and tie-breaking. Dedicated software was implemented to simulate the algorithm within the CERT-UA environment, enabling real-time registration, evaluation, and visualization of prioritized incidents. The simulation tested the algorithm’s effectiveness in handling incident inflows and its potential to streamline response efforts. Conclusions: this study presents a robust and novel multi-factor methodology that overcomes the insufficient granularity of existing national criticality levels. Introducing a hierarchical tie-breaking mechanism, the approach provides CERT-UA with a clear, decisive, and efficient tool for incident prioritization. Simulation and pilot implementations confirm the algorithm’s practical value and immediate applicability within the existing operational environments, significantly enhancing the ability of national-level response teams to mitigate the negative impacts of cyber threats. The system’s simplicity and adaptability ensure its applicability within existing operational environments, while its tie-breaking mechanism minimizes the risk of suboptimal prioritization. Future research directions include integrating artificial intelligence and machine learning to enhance prioritization accuracy and adapting this methodology for diverse organizational contexts. This work lays a strong foundation for advancing cyber incident management, addressing the evolving nature of cybersecurity challenges.
This study aims to determine the prerequisites for the occurrence of road traffic accidents, analyze the most dangerous maneuvers of motor vehicles that can lead to hazardous situations, and develop the most effective method for promptly informing the driver about potential danger. The goal of this study is to develop an information system that ensures timely notification of drivers about possible road traffic accidents in designated hazardous areas. The tasks include: investigating existing computer vision models for classification and object tracking tasks and determining the most suitable ones for deployment on a single-board computer Nvidia Jetson, while examining their performance and technical limitations; developing an optimized solution for the prompt notification of drivers about danger; creating an algorithm for detecting potential vehicle collisions that integrates computer vision methods and mathematical modeling; developing a comprehensive danger warning system based on the obtained results and testing its functionality. The following methods were applied in this study: a process-based approach to investigate the mechanisms of road traffic accident occurrence, statistical analysis of hazardous areas and maneuvers, and performance analysis of computer vision models for real-time object detection and tracking and driver notification. Additionally, road situations were simulated and modeled using the BeamNG.tech environment. The results include the development of a methodology based on computer vision and mathematical models for identifying hazardous situations on the road and the creation of an approach for prompt notification of road users using cloud technologies, IoT devices, and the GeoHash algorithm. An information system that allows drivers to receive warnings about potential hazards along their route has been proposed. Conclusions: this study confirms the successful development of a software system for forecasting and notifying drivers about the risk of road traffic accidents. The conducted studies have demonstrated the effectiveness of the proposed algorithm for detecting hazardous situations and technological solutions for road infrastructure integration. Experiments conducted using BeamNG.tech have confirmed the functionality of the developed system, which can be applied to minimize the risk of road traffic accidents in designated hazardous areas.
The subject matter of the article is the forecasting of time series of sea ice extent using statistical and deep learning methods. Sea ice extent is one of the most important indicators of climate change. Today, there are trends towards melting glaciers, which leads to a rise in sea level and, in turn, creates a significant threat of flooding of coastal regions around the globe. In addition, melting glaciers affect the flora and fauna of the Arctic and Antarctic regions, as well as economic stability in the world, covering economic development and food security. The spheres of agriculture, tourism, logistics are directly dependent on climate change, therefore, forecasting future changes is critically important for stability and sustainable development. The article analyzes the main trends in the change in sea ice extent. The goal of the study is to increase the reliability of long-term forecasting by designing a framework that covers the full forecasting cycle from data analysis to the use of predictive statistical methods and deep learning techniques. The tasks of the article are to conduct a comparative analysis of statistical methods and deep learning methods and their evaluation for the task of forecasting the area of sea ice distribution. The study used forecasting methods based on statistical models and deep learning. A study was conducted on the use of different approaches to forecasting future changes in a time series based on statistical methods, deep learning methods and ensemble models. The results obtained allow to evaluate the performance of models in the short term and an approach to long-term forecasting was formed. The use of autoregressors and deep learning methods is proposed to create a reliable long-term forecast. The comparison of the performance of the methods was carried out for the Northern and Southern Hemispheres. Conclusions. The scientific novelty of the results obtained is as follows: the method of forecasting time series of sea ice distribution using statistical methods and deep learning methods has been further developed. It was propose a generalizable forecasting framework that links time-series characteristics to model class selection and ensemble construction. The use of ensemble approaches allows us to ensure both the consideration of the main trends and the recognition of hidden patterns. The results obtained allow for a comprehensive assessment of time series for the Northern and Southern Hemispheres and indicate the feasibility of using both statistical forecasting methods for data with clearly defined patterns, such as the Arctic region, and deep learning methods to recognize hidden patterns observed in time series data for the Antarctic region.
The subject of the article is pixel-wise classification of Sentinel-2 satellite imagery represented as three-channel data mapped to the RGB color space for convenient visualization, with specific attention to the challenges posed by sensor noise and lossy compression artifacts typical for satellite data. The goal is to develop and validate a classification approach that maintains high accuracy under substantial noise and compression, by combining Local Binary Patterns (LBP) texture descriptors with pseudocolor features and employing an efficient ensemble classifier. The tasks to be addressed are: to design a compact feature representation that integrates LBP-based texture information with pseudocolor; to train and tune an XGBoost classifier on these features and compare its performance with baselines that rely on pseudocolor information alone and with simple neural network models; to assess robustness to noise and compression artifacts across a range of compression levels. The methods used include extraction of LBP descriptors to capture local texture patterns, construction of pseudocolor features from RGB-mapped Sentinel-2 channels, and concatenation of these descriptors into joint feature vectors. An XGBoost algorithm is employed to build the classification model. Model effectiveness is evaluated using the F1 score as the primary metric under varying noise and compression conditions. Visual inspection of the resulting classification maps is used to corroborate quantitative results and to analyze spatial consistency and error patterns. Conclusions. The scientific novelty of the results is as follows: for the first time in the context of Sentinel-2 pixel-wise classification, the use of LBP in combination with XGBoost has been systematically investigated and substantiated for BPG lossy compression scenarios at the optimal operating point (OOP) or nearby; it has been experimentally established that there is a substantial gain in classification accuracy for heterogeneous classes (urban areas, vegetation, bare soil) and a limited gain for homogeneous ones (water), and interaction artifacts of BPG+LBP on homogeneous surfaces have been documented, with directions outlined for adapting LBP parameters to mitigate them; the computational suitability of the approach (feature extraction, training, and classification time) for operational pipelines has been demonstrated; a comparison with a simple neural network has been conducted showing higher stability of the proposed approach on texture-rich classes under noise and compression, thereby delineating the limits of applicability of alternative methods. The study also shows that accounting for compression effects is important for operational processing pipelines: compressing images to an optimal operating point can reduce data volume and, in some cases, slightly improve classification accuracy by attenuating noise.
The object of research in this article is a combined architecture of analog-to-digital converters (ADCs), which is built by integrating a low-resolution flash ADC with a successive approximation register (SAR) ADC. Flash ADCs provide extremely high conversion speeds but suffer from a significant drawback: the resolution cost per bit increases exponentially with increasing bit depth. In contrast, SAR ADCs are characterized by low cost per bit, but their inherently sequential conversion mechanism limits their conversion speed. This study investigates a combined ADC architecture designed to effectively merge the advantages of flash and SAR ADCs, thereby maximizing economic efficiency per resolution bit. The core hypothesis is that using a flash ADC of relatively low resolution for initial rapid coarse conversion, followed by a SAR ADC for precise computation of the residual analog signal, can significantly reduce the overall cost of implementing high-resolution ADCs. The research objectives include analyzing the characteristics of flash and SAR ADCs, determining the optimal combination of their respective resolutions, developing the operational algorithm of the proposed combined ADC, creating a mathematical model in MATLAB Simulink, and evaluating its technical and economic performance. The results demonstrate that the optimal combination is a flash ADC with a resolution of 4–5 bits paired with an 8–10-bit SAR ADC. This configuration significantly lowers the cost per bit compared with traditional high-resolution flash ADCs while maintaining a considerably higher conversion speed compared with SAR ADCs of equivalent resolution. The simulation results indicated that integral nonlinearity (INL) and differential nonlinearity (DNL) values of the proposed ADC did not exceed ±0.5 LSB, confirming high conversion accuracy. In addition, we show that the energy-per-conversion figure remains unchanged relative to pure flash and pure SAR solutions in isolation. Furthermore, the economic analysis demonstrated that the proposed combined approach minimizes the implementation costs per unit of resolution. Conclusions. The proposed combined ADC architecture demonstrates substantial economic benefits compared with conventional flash ADCs and notably improved speed characteristics compared with SAR ADCs. The resolution distribution between flash and SAR components efficiently balances economic and technical requirements. Further studies should focus on the practical implementation of the proposed ADC architecture, noise impact analysis, and adaptive resolution management strategies.
The subject matter of this study is a collision checking for redundant robotic manipulators operating in variable environments, which remains a significant computational bottleneck in motion planning. The goal of this study is to improve computational efficiency of collision checking for multi-joint robotic manipulators in sampling-based motion planning, while preserving functional safety. This is achieved by developing and evaluating a learning-based method that predicts per-link penetration depth and serves as a statistical pre-filter rather than a replacement for exact collision checking. The tasks are as follows: 1) to propose a novel input representation that fuses the manipulator's kinematic state with localized geometric context extracted from the environment via voxel grids; 2) to design and implement a hybrid neural network architecture combining a fully-connected projection layer with a Kolmogorov-Arnold Network (KAN); 3) to train the network on a large, procedurally generated dataset of diverse collision scenarios; and 4) to evaluate the model's regression accuracy, classification performance, and computational speedup over a direct physics-based checker. The following results were obtained: the trained model achieves high regression accuracy with a low Mean Squared Error of 0.000148 on the test set; the model achieves promising classification results with a per-link recall of 93.01%, which is an important indicator for its use as a pre-filter capable of screening out the majority of hazardous states; computational speedup - performance benchmarks for a batch of 8192 states show that the proposed approach, including data preparation and inference, is approximately 3.7 times faster than a direct physics-based checker. Conclusions. The scientific novelty of results obtained is as follows: 1) a neural network architecture combining fully-connected and Kolmogorov–Arnold Network layers is proposed for predicting per-link collision depth of a redundant manipulator; 2) integration of kinematic and voxel-based geometric features into a unified input representation for accurate collision estimation. The proposed method effectively serves as a pre-filter for sampling-based planners, reducing the number of expensive collisions checks and accelerating the overall motion planning process.
The subject of this article is the process of estimating the parameters of a pulse signal with linear frequency modulation (LFM) used in airborne radar systems, particularly in synthetic aperture radars (SAR). The goal of this study is to synthesize algorithms for the optimal estimation of the key parameters of an LFM signal (i.e., carrier frequency, modulation frequency change rate, pulse length, and radio pulse envelope) and to develop a block diagram of a radar receiver that implements these algorithms. The tasks to be solved are as follows: build a mathematical model of a signal with linear frequency modulation emitted by a radar, an observation equation, and a likelihood functional; synthesize algorithms for estimating the parameters of an LFM signal using the maximum likelihood method; and develop a block diagram of a receiver based on the synthesized algorithms. The solutions to these tasks are based on the statistical theory of radio engineering systems and computer simulation. The following results were obtained: 1) algorithms for estimating the carrier frequency, frequency change rate, pulse length, and radio pulse envelope were synthesized; 2) simulations showed high noise robustness of the algorithms (up to a signal-to-noise ratio of –30 dB); 3) a block diagram of the radar was designed, which implements the synthesized algorithms and refines the estimated parameters in feedback. Conclusions. The scientific novelty of the obtained results is as follows: algorithms for estimating the parameters of both the point (carrier frequency, modulation frequency change rate, pulse length) and time characteristics (radio pulse envelope) of pulse LFM signals have been obtained using the maximum likelihood method. For the first time, it has been shown that estimating the pulse width requires solving a transcendental equation, and estimating the envelope requires smoothing in a sliding window. The obtained results expand the application of the maximum likelihood method in signal parameter estimation theory. The theory of phantomization of radio images has been further developed in terms of designing the receiving paths of phantomization radars.
This study focuses on the development and evaluation of an explainable artificial intelligence (XAI) framework for multimodal sentiment analysis, specifically applied to territorial revitalization project management. The research addresses the critical problem of “black box” AI models, whose lack of transparency hinders their adoption by project managers who require trustworthy information for high-stakes decision-making in complex social environments. The goal of this study is to propose and rigorously validate a novel framework for multimodal sentiment analysis that is tailored to provide transparent, trustworthy, and actionable insights for decision-making in territorial revitalization project management. The tasks to be solved include developing a hybrid XAI technique that fuses insights from cross-modal attention and gradient-based attribution, designing a cohesive, user-centric explanation format combining highlighted text and image heatmaps, constructing a custom RevitalizeSent-MM dataset for this specific domain, and empirically evaluating the framework’s predictive accuracy and, crucially, the fidelity of its explanations. The methods used involve a transformer-based Multimodal Sentiment Analysis (MSA) model using BERT and ViT with cross-modal attention for information fusion. The explainability component is a hybrid XAI technique that integrates cross-modal attention analysis with Integrated Gradients to assign importance scores to input features. Evaluation was performed using standard classification metrics for performance and the “Accuracy Drop on Perturbation” metric for explanation fidelity. The results confirmed the efficacy of the framework. The multimodal model demonstrated superior accuracy over unimodal baselines, and the proposed XAI method achieved significantly higher fidelity than naive explanation approaches, demonstrating its ability to accurately reflect the model’s internal reasoning. The scientific novelty lies in three areas: the development of a fused, hybrid XAI technique specifically for transformer-based multimodal models, creation of a unique, domain-specific dataset for revitalization analysis, and validation of a methodology for adapting advanced XAI to solve critical trust and adoption barriers, thereby confirming its practical significance in project management.
The subject of the study is Markov processes used for the formal description of the dynamics of states of unmanned vehicles controlled by an operator through augmented reality–based human–machine interfaces. Within the scope of the research, unmanned aerial and ground vehicles are considered as complex multi-state technical systems whose functioning is determined both by their technical characteristics and by the specific features of human interaction with the control interface. The aim of the study is to assess the impact of augmented reality–based human–machine interaction interfaces on the error-free decision-making of unmanned system operators, as well as on their responsiveness during control and reaction to changes in system states. The objectives of this study are to develop Markov models for the following scenarios: (a) without considering system failures and operator errors, with full recovery; (b) without considering failures, but allowing for operator errors, with full recovery; (c) considering system failures without operator errors, with full recovery; (d) considering both system failures and operator errors, with full recovery; and (e) without considering failures, allowing for operator errors, with the presence of a redundant unmanned system. The resulting Markov chains are intended to be used for modelling and subsequent comparison of the impact of different operating conditions on the system. As a result of the study, the following were obtained: (a) a classifier of states of unmanned aerial vehicles within a hazardous environment monitoring system based on the possible presence of failures, operator errors, and system redundancy; (b) Markov models for various system operation scenarios; and (c) simulation results of system operation based on the developed Markov models. Conclusion. The scientific novelty is as follows: a method for assessing the availability of monitoring systems with augmented reality–based human–machine interaction interfaces is proposed, which is based on single- and multi-fragment Markov models that take into account operator actions, partial failures, and the availability of reserve unmanned aerial vehicles. The proposed method quantitatively evaluates the impact of augmented reality not only on subjective indicators but also on the overall system availability and reliability indicators.
The subject of the study is the integration of artificial intelligence (AI) methods into software-defined networks (SDN) for adaptive control of access to limited resources within the infrastructure of Internet of Things (IoT) ecosystems. The goal of this work is to develop a model and architectural solution for a hybrid Cloud-SDN-IoT framework with embedded AI components, enabling the intelligent allocation of network and computing resources, and to experimentally validate the improvement of the fair distribution of a limited IoT resource across different traffic patterns in an emulation environment. The main tasks of the research are: 1) to analyze modern approaches to energy-efficient resource management and security in SDN-IoT networks; 2) to create the architecture of a hybrid Cloud-SDN-IoT framework that combines centralized SDN network control with the flexibility of cloud infrastructure; 3) to develop an experimental methodology using machine learning components to improve resource allocation and reduce load imbalance among competing clients; 4) to evaluate the system’s efficiency in relation to the stated objectives and the fair distribution of limited IoT resources by assessing the request distribution and the accuracy of detecting resource access violations. The paper proposes an improved three-layer SDN architecture model incorporating AI-based analytics: the IoT infrastructure layer, the SDN control layer, and the cloud application layer. The experimental part was implemented in a virtual Linux environment using Mininet and Ryu, where the trained AI model makes decisions about allocating the limited resource. The experimental results demonstrated that integrating the AI module into the SDN controller workflow increases the accuracy of detecting resource access violations, reduces load imbalance among clients, and improves the stability of real-time request distribution. Conclusions. The scientific novelty of the obtained results lies in the development of a reproducible hybrid Cloud-SDN-IoT architecture model that enables adaptive management of access to limited IoT node resources by combining centralized SDN control with AI-based predictive analytics. The AI-enabled control loop increased the average fairness accuracy of request distribution from 79.2% to 90.98%, an increase of 11.78 percentage points (14.87% relative), demonstrating improved proportional access to the limited IoT TokenServer API while preserving stable, real-time request regulation. The practical significance lies in the potential application of the proposed approach to optimize access to limited cloud services, APIs, energy resources, or IoT devices in smart city systems, healthcare, or industrial networks. Further research will focus on expanding the AI components with various machine learning models, forming new datasets, and conducting comparative evaluations of each model’s effectiveness in dynamic SDN-IoT resource management and reproduction under real-world conditions.
The subject matter of this article is the methodology for mathematical modeling of gas-dynamic processes in supersonic converging-diverging nozzles. This study aims to develop and substantiate a comparative validation methodology for assessing the accuracy of simplified analytical isentropic models against high-fidelity computational fluid dynamics (CFD) simulations for two-phase compressible flows involving secondary gas entrainment. The tasks to be solved are as follows: implementing a numerical model of the gas-particle flow within a supersonic nozzle using the finite volume method (ANSYS Fluent); performing analytical calculations of particle velocity and temperature using the one-dimensional isentropic model; and conducting a comparative analysis to identify systematic deviations and substantiate the applicability limits of the simplified analytical approach compared to the numerical solution. The methods used are: the study of gas dynamics of a two-phase flow was studied by numerical modeling using a modern computing package based on the finite volume method ANSYS Fluent, as well as conventional gas dynamics. The following results were obtained: two-phase CFD simulation (carried out in ANSYS Fluent) and a one-dimensional isentropic model were employed to analyze the behavior of nickel particles under varying gas stagnation temperatures (440 °C, 520 °C, 620 °C) and particle diameters (10 μm, 25 μm, 40 μm). The CFD results, which incorporate real gas dynamics, including turbulence, viscous effects, and particle–flow interactions, were compared with analytical results. The CFD results show significantly lower particle velocities (by 50 ± 7%) and higher temperatures (by 22 ± 7%) compared with the isentropic model, primarily due to the inclusion of thermal losses, boundary layer development, and secondary flow effects. The latter arises from the atmospheric entrainment of the carrier gas and powder into the divergent section of the nozzle. These factors disrupt the analytical approach’s ideal expansion, reducing the gas and particle velocities while increasing particle temperatures. Conclusions. The scientific contribution lies in substantiating that the classic isentropic model reaches its applicability limit for low-pressure cold spray nozzles with downstream injection. The correction coefficients derived from CFD data are proposed to refine analytical models. The practical significance lies in creating a basis for automated design algorithms for supersonic nozzles, enabling the derivation of correction coefficients for analytical equations in the future without the need for repetitive, computationally expensive simulations.
Russia’s full-scale invasion of Ukraine led to Europe’s largest and fastest displacement since World War II. Poland received the largest inflow. Rapid movement can affect COVID-19 spread and stress testing, reporting, and vaccination systems. Aim: To evaluate whether the invasion-related displacement coincided with short-term departures in Poland’s national COVID-19 cases and deaths using an intervention-anchored counterfactual forecasting approach learned from pre-invasion trends. Object of the study: Daily COVID-19 cases and deaths in Poland. Data come from the WHO dashboard, which spans May 2020 and centers on the 24 February 2022 invasion with a 30-day post-invasion horizon. Methods: We fit a univariate stacked LSTM to pre-invasion data and forecast one step ahead for the first 30 days after 24 February 2022. The network uses LSTM(128) - LSTM(64) - Dense(25) - Dense(1) with a linear output. The timeline is split into training (before 24 January 2022), validation (24 January-23 February 2022), and testing (24 February-+30 days). Each series (cases, deaths) fits six times with different random starts. Accuracy is measured using the mean absolute percentage error (MAPE). Deviations from the counterfactual are summarized as absolute and relative effects over the 30-day window. Results: Observed daily values closely tracked the counterfactual during the first month after the invasion, with only modest, short-lived over-prediction in the middle of the window. Between the validation and test periods, the average MAPE rose from 5.94% to 14.39% for cases and from 5.90% to 14.62% for deaths, reflecting greater short-run uncertainty but no large national-level break. Conclusion: Despite exceptional migration pressure, Poland’s national COVID-19 series did not show a marked divergence from a data-driven counterfactual in the first month after 24 February 2022. Scientific novelty: To the best of our knowledge, this study provides the first Poland-focused, short-horizon, data-driven counterfactual of the invasion shock under real-world Omicron conditions. It uses a simple, transparent LSTM trained only on pre-shock national data, repeats fits to capture training variability, and quantifies departures with clear absolute and relative measures.
This research aims to detect and identify unmanned aerial vehicles (UAVs) by analyzing the network traffic they transmit to their ground control station. The relevance of this topic arises from the need to develop highly efficient UAV identification systems, given their widespread use across the military, civilian, and commercial sectors. Effective solutions for detecting, identifying, and classifying these devices are required. This study aims to develop models and methods that enable machine-learning-based UAV identification systems by analyzing incoming and outgoing data traffic using an extremely intelligent information technology. A methodology is proposed for applying extreme information technology to UAV identification. This methodology involves employing pattern recognition techniques and designing intelligent information systems. The scientific novelty lies in introducing an approach to modeling an intelligent system capable of learning to identify UAVs based on traffic characteristics analysis. The research methodology includes the following steps: (1) forming a training dataset based on UAV traffic parameters, (2) building a UAV identification system using an information-extreme approach, and (3) training the system under standard operating conditions of digital radio communication channels used by UAVs. The results demonstrate that the proposed intelligent information system provides high UAV identification accuracy. Testing achieved an average identification accuracy of 86%. Conclusions. The proposed UAV identification system is based on an innovative approach to network traffic analysis using information-extreme intelligent technology. The results confirm its effectiveness for identification tasks under standard regulated traffic characteristics. The obtained results have practical significance for developing monitoring and protection systems in various fields against potential threats associated with UAV usage.
