2025

Постійне посилання на фонд

https://ela.kpi.ua/handle/123456789/78636

Переглянути

Зараз показуємо 1 - 20 з 29

Відкритий доступ
Approach to hybrid load management in Fat-Tree web clusters
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Radchenko, Kostiantyn; Chernenkyi, Artem
The paper presents an approach to hybrid load management in a web cluster that is capable of providing adaptive request balancing based on load prediction and resilience to random web server failures. The proposed architecture is built upon the Fat-Tree topology, which ensures high scalability, structural redundancy, and efficient routing within the cluster network. The developed system performs load forecasting using moving average methods and Erlang-based queueing models, enabling the estimation of overload probabilities and proactive redistribution of computational resources. Four representative simulation scenarios were analyzed: baseline load, peak load, dynamic traffic variations, and random server failures. The obtained results demonstrate enhanced system reliability, reduced average response time, and more balanced utilization of cluster resources. In the context of rapidly growing web services and user traffic volumes, the issue of maintaining high reliability and efficiency of clustered infrastructures becomes increasingly significant. Even with robust topologies such as Fat-Tree, irregular traffic patterns and sudden surges in client requests can cause local overloads and performance degradation. Random node failures further complicate cluster management, necessitating the use of adaptive and predictive control mechanisms. The proposed model integrates Fat-Tree network simulation with statistical forecasting algorithms, forming the basis for proactive load management. This integration allows for minimizing service degradation risks, dynamically responding to workload changes, and maintaining stable operation of web infrastructures under partial node failures. The architecture shows strong potential for real-time implementation in large-scale distributed web systems. It can be further enhanced by incorporating machine learning or wavelet-based forecasting methods to improve the accuracy of load estimation and system adaptability.
Відкритий доступ
Intelligent traffic management method in software-defined networks based on behavioral classification and adaptive priority service
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Oboznyi, Dmytro; Kulakov, Yurii
The growing complexity of modern enterprise network environments demands sophisticated traffic management solutions that can provide quality of service (QoS) guarantees for encrypted and heterogeneous flows. Existing traffic management approaches face significant challenges when dealing with encrypted protocols and diverse application requirements, resulting in performance degradation for critical services and inefficient resource utilization. This paper addresses the problem of intelligent traffic management in software-defined networks through behavioral classification and adaptive priority service mechanisms. The study examines the development and implementation of an integrated traffic management method that combines behavioral deep packet inspection, class-based queuing, and weighted random early detection algorithms. The research investigates how behavioral flow characteristics remain observable in encrypted traffic environments and how these patterns can be leveraged for effective QoS provisioning. The proposed method utilizes packet timing patterns, connection behaviors, and flow statistics to classify traffic without relying on payload inspection or predefined port assignments. Experimental validation through discrete-event simulation demonstrates significant performance improvements compared to traditional first-in-first-out mechanisms. The behavioral classification component achieves over 95% classification accuracy. The experimental results demonstrate up to 97.5% improvement in latency performance and 0% packet loss for high-priority traffic. Integrating behavioral traffic recognition with adaptive queue management within a programmable network framework provides an effective and innovative approach to maintaining stable service quality in encrypted, multi-service environments. The proposed method is compatible with existing software-defined network controllers and can be deployed without modification of application protocols or infrastructure components.
Відкритий доступ
UAeroNet: domain-specific dataset for automation of unmanned aerial vehicles
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Kochura, Yuriy; Trochun, Yevhenii; Taran, Vladyslav; Gordienko, Yuri; Rokovyi, Oleksandr; Stirenko, Sergii
This paper addresses the challenges and key principles of designing domain-specific datasets that canbe used especially for automation of unmanned aerial vehicles. Such datasets play a key role in buildingintelligent systems that enable autonomous operation and support data-driven decisions. The study presentsapproaches we used for data collection, analysis and annotation, highlighting their importance and practicalimpact on real-world application. The preparation of a domain-specific dataset for automating unmannedaerial vehicles operations (such as navigation and environmental monitoring) is a challenging task due tofrequently low image resolution, complex weather conditions, a wide range of object scales, backgroundnoise and heterogeneous terrain landscapes. Existing open datasets typically cover only a limited variety ofunmanned aerial vehicles use cases, which restricts the ability of deep learning models to perform adequatelyunder non-standard or unpredictable conditions.The object of the study is video data acquired by unmanned aerial vehicles for creating domain-specificdatasets that enable machine learning models to perform autonomous object recognition, navigation, obstacleavoidance and interaction with an environment with minimal operator involvement. The subject focuseson the collection, preparation and annotation of video data acquired by unmanned aerial vehicles. Thepurpose of the study is to develop and systematize workflow for creating specialized datasets to trainrobust models capable of autonomously recognizing objects in real-time video captured by unmanned aerialvehicles. To achieve this goal, a workflow was designed for collecting and annotating video data, raw videodata were acquired from unmanned aerial vehicles sensors and manually annotated using the ComputerVision Annotation Tool.As a result of this work, we developed a domain-specific dataset (UAeroNet) using an open-sourceannotation tool for object tracking task in real scenarios.UAeroNetconsists of 456 annotated tracks and atotal of 131 525 labeled instances that belong to 13 distinct classes.
Відкритий доступ
Evaluation of the effectiveness of two approaches to building damage detection with satellite imagery
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Oliinyk, Yurii; Rumiantsev, Oleksii
This study addresses the approaches for satellite image analysis to assess infrastructure damage. Themain aim is to conduct a comprehensive comparative analysis of the effectiveness of two key machinelearning approaches: specialized semantic segmentation based on theU-Netarchitecture and generalizedvisual analysis using large vision-language models. The object of the research is the process of quantitativelybenchmarking these two distinct approaches to determine their practical applicability for multi-class damageclassification.The research material is the publicly availablexView2dataset. The methods involved two parallelexperiments. For the semantic segmentation approach, aU-Netmodel with anEfficientNet-B4encoderwas implemented and trained on 6-channel input data (”before” and ”after” images) using a combinedDiceandFocalloss function. For the vision-language models approach, the open-sourceLLaVA-1.5-7Bmodelwas evaluated in a zero-shot mode using advanced prompt engineering for an aggregative counting task.To enable a direct comparison, the standardJaccard indexwas calculated based on the aggregated objectcounts for each damage class.The results of the experiments revealed a significant performance disparity. The specializedU-Netmodeldemonstrated high effectiveness, achieving an intersection over union score of 0.6141 on the test set. Incontrast, theLLaVAmodel proved unsuitable for accurate quantitative analysis, yielding an extremely lowJaccard indexof approximately 0.063, primarily due to its systemic failure to correctly identify and countobjects (𝑅𝑒𝑐𝑎𝑙𝑙≈0.07). The scientific novelty lies in being the first study to quantitatively document thisorder-of-magnitude capability gap, confirming that for tasks requiring high-precision mapping, specializedsegmentation models remain the indispensable tool.
Відкритий доступ
DDOS attack detection with data imperfections using machine learning algorithms
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Dremov, Artem; Volokyta, Artem
The issue of DDoS attacks remains a prevalent one even in recent years. Modern environment is highlydynamic and is characterized by a large amount of traffic flow. Existing research covers several models,techniques and approaches to detecting DDoS traffic, which aim to optimize the detection in controlleddatasets. However, unintentional noise or data corruption may lower the efficacy of such methods. As such,determining most effective ways to detect DDoS traffic in conditions of data imperfections is necessary forreliable network performance.Therefore, the object of this research Is the usage of machine learning algorithms for detection ofincoming DDoS attacks. The purpose of this research is to determine the performance of ways to detectincoming DDoS attacks with machine learning algorithms based on detection accuracy, while simulatingimperfect data conditions. The study also examines the impact of class rebalancing on modified data.To achieve the aim of this research a variety of machine learning algorithms were implemented andtested on aCIC-DDoS2019dataset. The data is modified by removing values and introducing noise, tested,the classes are resampled and the dataset is tested again. The goal is to achieve over 90% accuracy in aclassification task of the type of DDoS attack and to determine how much the changes affect the performanceof the algorithms.The results of the testing indicated that several solutions reach the target mark and changes to thedataset in realistic conditions do not significantly affect the final result. However, all models tested showa decrease in accuracy compared to unmodified data with more complex models showing higher resilience(smaller decrease in accuracy). In addition, resampling of the data shows comparable decrease in accuracyof the models with more complex models being affected less.The results of this study may be used in development of an algorithm of repairing the corrupted dataor development of models more resistant to such data changes. Additionally, the results of this study maybe used when considering models for practical implementations of a DDoS traffic classification system.
Відкритий доступ
Optimization neural network for time series processing
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Pysarchuk, Oleksii; Baran, Danylo
The article proposes the architecture of the optimization neural network and the model of test samplesynthesis for the process of extrapolation of time series parameters. In particular, the addition of an inputlayer with the introduction of an optimization scheme of nonlinear trade-offs has been implemented.Extrapolation of the behavior of the time series was carried out according to a test sample, which isformed as a data model with the selection of the trend according to the method of least squares. Thescientific novelty of the results obtained in the article is reflected in the essence of these decisions.The aim of the research is to develop an optimization network architecture and data model forextrapolation, which allows to improve the accuracy and time of predicting the behavior of the time seriesoutside the observation interval. Subject of research: architecture of an artificial neural network andmethods of extrapolation of time series. Object of research: processes of architectural synthesis of anartificial neural network and extrapolation of time series behavior outside the observation interval.The optimization layer provides mini-requirements for the approximation of training and test samples.This is especially appropriate for time series with stochastic noise and allows you to reduce the impactof random errors on time series prediction results. The use of model data for extrapolation allows you todetermine the behavior of the time series outside the observation interval. At the same time, the forecastingtime with acceptable accuracy characteristics increases. These solutions are reflected in the name of theoptimization neural network, which is proposed by the authors. The study of the effectiveness of the proposedsolutions was implemented by methods of simulation modeling on a modified artificial neural network. Theresults of the calculations proved an increase in the adequacy of data models and an increase in the accuracyof extrapolation.
Відкритий доступ
Comparative analysis of LCNet050 and MobileNetV3 architectures in hybrid quantum–classical neural networks for image classification
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Khmelnytskyi, Arsenii; Gordienko, Yuri
This study explores the impact of classical backbone architecture on the performance of hybrid quantum-classical neural networks in image classification tasks. Hybrid models combine the representational power of classical deep learning with the potential advantages of quantum computation. Specifically, this research employs a quanvolutional neural network architecture in which a quantum convolutional layer, based on a four-qubit Ry circuit, preprocesses input images before classical processing. Despite the growing interest in hybrid models, few studies have systematically investigated how variations in classical architecture design affect the overall performance of hybrid quantum-classical neural networks. To address this gap, we compare two lightweight convolutional backbones – MobileNetV3Small050 and LCNet050 – integrated with an identical quantum preprocessing layer. Both models are evaluated on the CIFAR-10 dataset using 5-fold stratified cross-validation. Performance is assessed using multiple metrics, including accuracy, macro- and micro-averaged area under the curve, and class-wise confusion matrices. The results indicate that the LCNet-based hybrid model consistently outperforms its MobileNet counterpart, achieving higher overall accuracy and area under the curve scores, along with improved class balance and robustness in distinguishing less-represented classes. These findings underscore the critical role of classical backbone selection in hybrid quantum-classical architectures. While the quantum layer remains fixed, the synergy between quantum preprocessing and classical feature extraction significantly affects model performance. This study contributes to a growing body of work on quantum-enhanced learning systems by demonstrating the importance of classical design choices. Future research may extend these insights to alternative datasets, deeper or transformer-based backbones, and more expressive quantum circuits.
Відкритий доступ
Detection Method of Fraudulent Payment Transaction Based on C-Score Metric
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Korynetskyi, Dmytro; Stetsenko, Inna V.
Fraud detection for payment transactions is a cost-sensitive task, as the costs associated with misclassification – such as missing a fraudulent transaction or incorrectly blocking a legitimate one – can vary significantly depending on business priorities. Traditional evaluation metrics, particularly the F1-score, ignore this asymmetry, creating a need for more flexible approaches. This research focuses on developing a method for building adaptive, cost-sensitive fraud detection systems. The aim is to develop a method that enables the practical application of the cost-sensitive C-score metric to configure a multi-level decision logic. The paper also presents a possible software architecture for its implementation. The proposed two-phase method (offline calibration and online scoring) uses the C-score metric to determine multiple decision thresholds corresponding to different business scenarios. Its validation was conducted on the public “Credit Card Fraud Detection” dataset using the XGBoost algorithm. The Synthetic Minority Over-sampling Technique (SMOTE) was applied to overcome the severe class imbalance in the data, and a comparison was made against the traditional F1-score-based approach. The experimental results showed that the proposed approach allows for the identification of two distinct thresholds from a single classifier. The first threshold ensures high precision, making it suitable for automated blocking of payment transactions with minimal false positives. The second threshold, focused on high recall, enables the selection of suspicious payment transactions for subsequent manual review. It was also confirmed that the SMOTE significantly contributed the model's class separation ability, thereby increasing the reliability of calibrating these thresholds. Based on the method, a practical blueprint for a service-oriented architecture is proposed for creating flexible and configurable anti-fraud systems.
Відкритий доступ
Automatic Network Reconfiguration Method with Dynamic IP Address Management
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Haidai, Anatolii; Klymenko, Iryna
In the context of growing cyber threats, systems capable not only of detecting anomalies in the operation of network infrastructure but also of promptly responding to them without administrator intervention are becoming increasingly relevant. This paper proposes a method for automatic network reconfiguration based on the integration of the Zabbixmonitoring system with the pfSensenetwork gateway functionality. Such a system enables centralized control of the operating system status, resource usage, and network activity, while also allowing for automatic changes to host IP addresses, routing adaptation, and connection restrictions according to defined security policies.The aim of the study is to develop a method for automatic network monitoring and reconfiguration with dynamic IP address changes to improve the effectiveness of cyber threat mitigation. The object of the study is the processes of information security management in computer networks. The subject of the study includes methods of anomaly detection and automatic response through modification of network parameters using Zabbixand pfSense.In the context of automatic response to detected threats, the method of comprehensive monitoring of client host operating systems has been formalized, including subsequent analysis of logs, user actions, resource load, network port usage, and interaction with external services. A methodology for network reconfiguration after anomaly detection has been developed and implemented: in particular, changing the IP address while maintaining functionality in a minimal network access configuration and isolating the node using pfSense. Scripts for Windows client OS were employed, interacting with the Zabbixand pfSenseAPIs, thus ensuring dynamic and fully automated operation.Testing results of the proposed system in a simulated environment confirm its effectiveness. Compared to manual or partially automated solutions, incident response time was reduced, and the risk of attack propagation within the network was minimized.
Відкритий доступ
Vision-Based Neighbor Selection Method for Occlusion-Resilient Uncrewed Aerial Vehicle Swarm Coordination in Three-Dimensional Environments
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Smovzhenko, Oleksii; Pysarenko, Andrii
Uncrewed aerial vehicle (UAV) swarms provide superior scalability, reliability, and efficiency compared to individual UAVs, enabling transformative applications in search and rescue, precision agriculture, environmental monitoring, and urban surveillance. However, their dependence on Global Navigation Satellite Systems (GNSS) and wireless communication introduces vulnerabilities like signal loss, jamming, and scalability constraints, particularly in GNSS-denied environments. This study advances swarm robotics by developing a novel neighbor selection method for occlusion-resilient, vision-based coordination of UAV swarms in three-dimensional (3D) environments, addressing the problem of visual occlusions that disrupt decentralized flocking. Unlike prior research focusing on planar settings or communication-dependent systems, we model swarm coordination as an artificial potential field problem. Additionally, we evaluate performance through metrics like minimum nearest neighbor distance (collision avoidance), alignment (velocity synchronization), and union (cohesion). Using simulations in point mass and realistic quadcopter dynamics (Gazebo with PX4) environments, we assess swarm behavior across dense, default, and sparse configurations. Our findings reveal that occlusions degrade alignment (below 0.9) and distances (below 0.5 m) in dense swarms exceeding 70 agents, increasing collision risks. Our novel method, incorporating metric, topographic, and Delaunay strategies, mitigates these effects. Topographic selection achieves high alignment (above 0.9) in small swarms (up to 50 agents), while Delaunay ensures perfect cohesion (union = 1) and robust alignment across all swarm sizes. Validation in simulations confirms these results. Furthermore, our method enables communication-free coordination that matches or surpasses communication-enabled performance, with topographic selection outperforming (alignment above 0.9 vs. 0.85) in small swarms and Delaunay excelling in larger ones. This result eliminates the need for inter-agent communication, enhancing resilience and bandwidth efficiency. These findings establish a scalable, infrastructure-independent framework for UAV swarms, with practical value for autonomous operations in complex, occlusion-prone environments.
Відкритий доступ
Hybrid Voting Model for Decentralized Autonomous Organizations with Dynamic Quorum
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Serebriakov, Roman; Klymenko, Iryna
The article examines the problem of balancing security and flexibility in decision-making mechanisms within decentralized autonomous organizations (DAOs), which operate without centralized control through the use of smart contracts. To this end, two main voting models employed in DAOs are analyzed: the conjunctive model, which requires unanimous approval of a proposal by all participant groups, and the disjunctive model, where approval from a single group is sufficient. Both models have significant advantages and drawbacks: the former ensures a high level of security and protection of all parties’ interests but considerably slows down the decision-making process, while the latter provides speed and scalability but introduces risks of centralized influence. In response to these challenges, a hybrid voting model is proposed, in which the type of logic is determined by the nature of the proposal. Specifically, critical changes, such as updates to governance rules or quorum parameters, must involve all groups, whereas routine operational matters can be decided through a simplified disjunctive procedure. The implemented smart contract architecture supports both mechanisms and enables DAOs to dynamically adjust quorum thresholds through separate governance proposals. To evaluate the effectiveness of the model, a simulation of 1,000 voting processes was conducted under four different scenarios of participant activity: balanced, one-sided, and low overall participation. The results showed a reduction in the probability of deadlock situations and an increase in the share of successful votes when hybrid logic was applied, particularly under conditions of low or asymmetric participation. In addition, special attention was given to gas cost optimization: the disjunctive approach allows vote counting to be stopped once a quorum is reached by one group, thus reducing overall computational expenses. Therefore, the proposed solution appears promising for both financial DAOs and decentralized infrastructures, particularly the Internet of Things, where speed, scalability, and secure coordination are especially important.
Відкритий доступ
Hybrid Path Planning Method for Unmanned Ground Vehicles Swarm in Dynamic Environments
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Rudnytskyi, Myroslav; Klymenko, Iryna
Unmanned ground vehicles (UGVs) have significant potential across various applications. These include automation of the agricultural tasks, inspection and maintenance within construction and industrial sectors, automation of complex assembly processes and infrastructure repairs, explosives disposal, automation of logistical operations, search-and-rescue missions, and expeditions to hard-to-reach or hazardous areas. However, a key challenge limiting their widespread deployment is autonomous navigation, which remains a significant problem due to dynamic environments characterized by constantly changing obstacle configurations, unpredictable scenarios, and the need for rapid real-time decision-making to ensure safe and stable movement. The object of this paper is a hybrid path planning for the autonomous navigation of unmanned ground vehicles swarm within a simulated environment. The research aims to develop autonomous navigation method for the unmanned ground vehicles swarm by employing a hybrid approach designed to enhance the efficiency of obstacle avoidance and improve the adaptability to dynamic environments. To achieve this goal, a novel autonomous swarm navigation method based on a hybrid approach is proposed. This approach differs from existing solutions by employing the A* path planning algorithm with incorporated traversal costs on the map for global-level navigation and the artificial potential field (APF) algorithm, that supports linear and V-shaped formations for local-level navigation. The research findings indicate that the proposed method allows the swarm to perform optimal path planning, considering traversal costs, and effectively avoid local minimum problems that are inherent to the artificial potential field method. The successful performance of the method within the simulated environment demonstrates its potential for future validation in real-world scenarios and practical applications involving swarms of unmanned ground vehicles operating in challenging environments. At the same time, the study identified challenges related to swarm size scalability in narrow spaces, defining directions for further improvements.
Відкритий доступ
Statistical Evaluation of Parameters in Nonlinear Models Using Integral-Form of Least Squares Method and Differential Non-Taylor Transformations
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Pysarchuk, Oleksii; Tuhanskykh, Oleksandr
The article presents the method for statistical leaning of nonlinear model parameters, its principle and efficiency of application. It proposes a solution to the shortcomings of using the differential spectra balance approach through the integral form of least squares in the scheme of differential non-Taylor transformations. The object of the study is the process of statistical learning of nonlinear model parameters. The purpose of this paper is to formulate the statistical learning method for evaluation of nonlinear model parameters using LSM in integral form and differential non-Taylor transformations. It is relevant in many areas of modern activity and is necessary for applying the statistical training methodology to a more complex time series, as well as increasing the accuracy of the received expectations. To achieve this goal the statistical learning methodology was proposed, which is based on the creation of process model in integral form of least squares with simplification using non-Taylor transformations. It differs from existing approaches by incorporating all the available differential discretes in the created model, which allows for better predictability and circumvents the problem of unequal count of discretes inside the models, which allows for better application of the method for different model forms. The algorithm for the process was formed, using which it can be applied different models. In the paper several experiments were conducted to verify the efficacy of the proposed method in different situations. These experiments use generated datasets that are polluted with stochastic errors to better simulate real data. The results of modeling are shown, and the statistical characteristics of the obtained expectation are compared with the results of the application of the statistical training methodology using the differential spectra balance. During the study, features of the technique were found that must be considered for its application to data with a high number of stochastic deviations. Based on the obtained metrics, a conclusion is made about the effectiveness of using the technique relative to time series reflecting processes of different nature.
Відкритий доступ
Drone Swarm Control Model Based on High-Level Petri Nets
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Ivankov, Valentyn; Novotarskyi, Mykhailo
The rapid growth of unmanned aerial vehicle (UAV) applications in the modern world imposes significant demands on the reliability of control logic. An error in the sequence of stages can lead at best to inefficient battery usage or violations of airspace regulations, and at worst to an accident with loss of the vehicle and potential harm. Control is usually implemented using scripts or behavior trees, which complicates maintenance. The reason is that the size of the source files quickly increases, and when it becomes necessary to add new functionality or modify existing logic, there is a risk of introducing vulnerabilities by failing to account for all possible situations. This is why High-Level Petri Nets (HLPN) were chosen, as this method addresses the problem of formally describing the control system and allows the system to be easily scaled or modified in any way. The aim of the study is to develop and validate a model based on HLPN that will serve as the single source of truth for UAV swarm control. In the proposed model, the places correspond to flight stages, and the tokens carry numerical parameters such as battery charge, coordinates, and telemetry. Thus, a single scheme simultaneously describes discrete events and constraints. For each transition, conditions are formalized to verify the possibility of its execution, such as checking the minimum required battery level or verifying location. The methodology includes several stages. First, the network structure is formally defined. Then, based on this structure, a Python model is built that implements the developed network, controls movement between states, and ensures the correct sequence of transition firings. After developing the model, testing and analysis of the obtained results are performed. The results show that using HLPNs to build a model for verifying commands in a discrete mode indeed ensures a correct description of transitions between states and increases the reliability and survivability of the developed control system model, while also significantly reducing maintenance efforts. The developed model is easily adaptable to route changes, addition of sensors, or functional expansion.
Відкритий доступ
Hexacopter-Based Cyber-Physical System for Water Sampling with Adaptive Path Planning and Multi-Drone Coordination
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Pysarenko, Andrii; Rolik, Oleksandr
The object of this study is a hexacopter-based cyber-physical system designed for autonomous water sampling to support environmental monitoring, addressing the problem of inefficient control under dynamic conditions. The subject focuses on integrating physical flight control and water sampling operations with cyber supervisory functions, including real-time waypoint navigation, task scheduling, and multi-drone coordination, validated as a current system component. The research investigates the system’s performance under payload variations and wind disturbances, ensuring robustness and precision in adverse environments. The purpose is to improve efficiency of water sampling through this CPS, achieving enhanced flight stability and positioning accuracy via a cascade PID control system, optimizing mission planning with adaptive cyber strategies, and increasing scalability through multi-drone operations. This approach aims to surpass traditional UAV systems by using physical-cyber integration for precise, robust, and scalable water quality assessment. The methodology combines simulation-based and analytical techniques to develop and assess the hexacopter CPS. A 6-degree-of-freedom mathematical model, based on Newton-Euler equations, was constructed in MATLAB/Simulink to simulate hexacopter dynamics, incorporating payload and wind effects. The cascade PID control system was tuned using the Ziegler-Nichols method, with iterative optimization to reduce overshoot and settling time across three scenarios: 1 kg static payload, 1.5 kg dynamic payload, and 5 m/s wind. The cyber supervisory system, implemented in ROS 2, employs graph-based algorithms (Dijkstra’s for waypoint navigation, list-scheduling for task allocation) and a consensus protocol for multi-drone coordination, tested in a 500x500 m² environment. Performance metrics, such as position root mean square error (RMSE) and attitude errors, were analyzed to evaluate system effectiveness. Results demonstrate significant improvements in water sampling capabilities. The cascade control system achieved a 40–50% reduction in position RMSE and maintained attitude errors within ±0.8° to ±1.2°, ensuring stable flight. The cyber-physical framework reduced mission time by 15% through adaptive path optimization, while multi-drone coordination increased sampling coverage by 20%, enhancing scalability. These outcomes reflect the system’s precision and robustness that highlight novel control and coordination strategies with practical value for environmental monitoring. The study provides a foundation for future ecological applications.
Відкритий доступ
Decentralized Task Allocation Method in Hierarchical IoT Systems Using Fuzzy Logic
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Rolik, Oleksandr; Nahaiko, Dmytro
The use of fog and edge computing extends the computational capabilities of IoT systems to the network edge, contributing to the minimization of delays during task execution. Osmotic computing complements distributed computing by providing seamless integration between computational environments through dynamic migration of micro-elements across different hierarchy tiers according to current load conditions and resource availability. However, within the concept of osmotic computing, a key challenge remains the effective management of task allocation under conditions of uncertainty, dynamism, and heterogeneity of the IoT environment. The aim of this study is to improve the efficiency of resource utilization and task allocation in hierarchical IoT systems based on osmotic computing under uncertain and dynamically changing environmental conditions. The object of the study is the process of task allocation in multi-tier IoT systems that include cloud, fog, and edge computing. The subject of the study is methods and models for task allocation and computing resource management in IoT systems using the osmotic computing paradigm. The paper presents a three-tier hierarchical management model built on cloud, fog, and edge environments, which implements a centralized-decentralized management approach. Each tier is represented by a set of computing nodes and a management system that performs local task allocation, resource state monitoring, and micro-element management. The management system of the lower tier is subordinate to the higher-tier management system in the hierarchy. A method for decentralized task allocation in hierarchical IoT systems using fuzzy logic has been developed. The allocation method includes two decision-making stages using a fuzzy inference system: determining the direction of task allocation and selecting the optimal computing node for its execution. The determination of task allocation direction is carried out based on task characteristics, and the suitability rating of computing nodes is determined considering task execution latency, resource utilization efficiency, and load balancing. The task is assigned to the node with the maximum rating. The use of fuzzy logic ensures rational decision-making under conditions of uncertainty in real-time, which is characteristic of highly heterogeneous and dynamic IoT environments. Experimental modeling and investigation of the method were carried out using the iFogSim simulation environment. The research results show that the percentage of locally executed tasks remains virtually unchanged with different numbers of tasks, indicating stability in decision-making. Increasing the intensity of task generation leads to an increase in task computation latency due to increased load on computing nodes, while task assignment latency and response latency remain unchanged. The method demonstrated adaptability in task allocation for different types of tasks.
Відкритий доступ
Method for Software Pipelining on Graphical Processing Units
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Vinokurov, Artemii; Sergiyenko, Anatoliy
Graphics Processing Units (GPUs) play a significant role in high-end computations, including artificial intelligence. However, the GPU hardware is often underloaded. This forces an increased volume of GPU hardware to maintain a high throughput for task execution. The low loading of the GPU resources remains an actual problem and it needs to be solved now. Therefore, it is essential to seek methods that enhance GPU loading. The research object is computational processes in modern processors, especially in GPUs. The purpose of this study is to review the software pipelining approach, its advantages and disadvantages, the techniques that can be used in it, including both instruction-level and decoupled versions, and to assess the effectiveness of this approach for the GPU. To satisfy the requirements, different analysis methods were used. First, the architectural requirements to apply software pipelining were reviewed. Second, the original formulation and historical development of the approach were examined. Third, different levels of parallelisation to implement software pipelining were explored. Finally, C-slowing was proposed as an optimisation technique to overcome the adversities of the underutilisation of computational resources. The research has revealed the abundance of proper software pipelining for GPU implementations. Whereas existing works review the possibilities of this technique, they are often overlooked in contrast to simpler multi-threading techniques. However, investigated researchers have defined the crucial limiting factor to computational resources as a constraint by memory overloading, specifically the pipelining registers. To address this, the C-slowing approach was suggested and theoretically evaluated. It demonstrated a possible increase of over 30% in GPU loading for the analysed algorithm, proving its applicability. In conclusion, the software pipelining approach shows decent potential to optimise GPU algorithms, requiring further investigation. C-slowing could be utilised to handle the problem of underutilisation of computation.
Відкритий доступ
Method for On-line Acceleration of Dependent Operation Chains Using Redundant Code on FPGA with System of Linear Equations Example
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Verbovskyi, Illya; Zhabin, Valerii
This study examines methods for accelerating the execution of dependent operation chains in on-line mode through parallel processing of operands at the bit level in redundant code on field-programmable gate arrays (FPGA). The object of research is the hardware implementation of the Thomas algorithm for solving systems of linear equations with tridiagonal matrices on FPGA platforms. The aim is to develop a method for accelerating dependent operation chains in on-line mode using redundant code with minimization of pin count requirements. The methodology employs algorithmic analysis, hardware modeling using Active HDL, performance evaluation based on timing characteristics and resource utilization on Altera Cyclone III EP3C5E144 platform, with verification performed using Quartus. The results reveal bottlenecks in traditional FPGA implementations of the Thomas algorithm and demonstrate that the proposed optimized method provides over threefold performance improvement while maintaining constant pin count regardless of operand bit depth. The developed computing module architecture enables bit-wise parallel data processing and supports a modified version of the Thomas algorithm adapted for on-line operation. The scientific novelty lies in combining redundant code with on-line computation techniques to simultaneously achieve computational acceleration and hardware implementation simplification. The practical value is determined by the applicability of the proposed approach to resource-constrained FPGA platforms, ensuring efficient implementation of computationally intensive algorithms with dependent operation chains.
Відкритий доступ
Mathematical Model of Clustering of Informational Messages with Indicators of Activity for the Information Content by Tone and Areas of Society Activity
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Pysarchuk, Oleksii; Baran, Danylo
The mathematicalmodel of clustering of information messages has been further developed, which is based on the frequency analysis of their tonality using Natural Language Processing methodologies with the support of large language models; OLAP visualization of clustering results and is distinguished by an established system of indicators of information content activity by areas of society activity with hierarchical compression of incoming Big Data arrays, which determines the database model for their storage. This provides an improvement to the analysis of information messages in global information networks by taking into account many factors in the areas of society activity.The main idea and goal of the mathematical model for clustering information messages is to implement a sequence of preparation stages for detecting critical activity of the information content in global media. In practice, this is the establishment of a list and the determination of indicator values that measure content activity in primary messages, followed by their transformation into a time series –a systematized dataset. In the conditions of high density of the flow of occurrence, dynamics of development, and transformation of information content, a Big Data structure of information messages is taken into account. Therefore, the clustering model, apart from division by informational features, should provide the hierarchical compression of incoming Big Data arrays.Research objective: development of a mathematical model of clustering information messages with indicators of information content activity by tone and spheres of activity of society.Research subject: methods of clustering information messages.Research object: process of clustering information messages.
Відкритий доступ
A Multimodal Retrieval-Augmented Generation System with ReAct Agent Logic for Multi-Hop Reasoning
(National Technical University of Ukraine "Igor Sikorsky Kyiv Polytechnic Institute", 2025) Yuvzhenko, Denys; Chymshyr, Viacheslaw; Shymkovych, Volodymyr; Znova, Kyrylo; Nowakowski, Grzegorz; Telenyk, Sergii
The rapid advancement of generative artificial intelligence models significantly influences modern methods of information processing and user interactions with information systems. One of the promising areas in this domain is Retrieval-Augmented Generation (RAG), which combines generative models with information retrieval methods to enhance the accuracy and relevance of responses. However, most existing RAG systems primarily focus on textual data, which does not meet contemporary needs for multimodal information processing (text, images, tables). The research object of this work is a multimodal RAG system based on ReAct agent logic, capable of multi-hop reasoning. The main emphasis is placed on integrating textual, graphical, and tabular information to generate accurate, complete, and relevant responses. The system's implementation utilized the ChromaDB vector storage, the OpenAI embedding generation model (text-embedding-ada-002), and the GPT-4 language model. The purpose of the study is the development, deployment, and empirical evaluation of the proposed multimodal RAG system based on the ReAct agent approach, capable of effectively integrating diverse knowledge sources into a unified informational context. The experimental evaluation utilized the Global Tuberculosis Report 2024 by the World Health Organization, containing various textual, graphical, and tabular data. A specialized test set of 50 queries (30 textual, 10 tabular, 10 graphical) was created for empirical analysis, allowing comprehensive testing of all aspects of multimodal integration. The research employed methods such as semantic vector search, multi-hop agent-based planning with ReAct logic, and evaluations of answer accuracy, answer recall, and response latency. Additionally, an analysis of response speed dependence on query volume was conducted. The obtained results confirmed the high efficiency of the proposed approach. The system demonstrated an answer accuracy of 92%, answer recall of 89%, and ensured complete (100%) coverage of all data types. The average response time was approximately 5 seconds, meeting interactive system requirements. Optimal parameters were experimentally determined (for example, parameter k = 6, classification threshold 0.35, and up to three reasoning iterations), ensuring the best balance among completeness, speed, and operational efficiency. The study's findings highlighted significant advantages of the multimodal agent-based approach compared to traditional textual RAG solutions, confirming the promising direction for further research.

Переглянути

Перегляд 2025 за Дата публікації

Результатів на сторінці

Налаштування сортування