Перегляд за Автор "Besedovskyi, Oleksii"

Зараз показуємо 1 - 2 з 2
  • Ескіз
    Документ
    Decision-making method in interdependent computing systems
    (Хмельницький національний університет, 2025) Kryzhanyvskyi, Dmytro; Drozd, Andriy; Besedovskyi, Oleksii
    The relevance of this paper lies in the fact that modern interdependent computing systems are being actively implemented in critical areas ranging from smart energy grids and transportation systems to autonomous robotic platforms and distributed cloud services. These systems are characterized by a complex structure, a large number of interacting agents, and high requirements for real-time decision-making. Despite significant scientific and technological progress, a number of challenges remain unresolved to ensure the sustainability, adaptability, and coherence of all system components. One of the key challenges is the need to ensure rational decision-making in a decentralized environment where each agent has limited information about the state of the system as a whole and operates under conditions of uncertainty and potential distrust of other agents. Classical centralized methods are often ineffective or inapplicable in such cases due to excessive complexity or delays in data exchange. The issue of developing methods that ensure not only the correctness of decisions but also compliance with time constraints is particularly relevant. In interdependent computing environments, where the decision of one agent affects the outcome of the work of others, any delay or error in the strategy can lead to degradation of the performance of the entire system. In such environments, it is crucial to use adaptive, game-based, and reputation-based approaches that allow for dynamic consistency and stability of the system. In this paper, we develop a decision-making method for interdependent computing systems that combines Bayesian reputation updating, log-linear strategy learning, and reinforcement learning mechanisms. The peculiarity of the proposed method is its ability to adapt to changes in the environment and effectively detect unscrupulous agents by dynamically adjusting reputations. The algorithmic implementation of the model allows achieving the Bayesian-Nash equilibrium, which indicates the stability of the system even in complex interaction scenarios. The results of experimental modeling have demonstrated that the proposed method strikes a balance between adaptability, reliability, and efficiency of interactions. The system demonstrates the ability to self-organize, stabilizes in fewer iterations compared to classical approaches, and effectively prevents the influence of sabotaging behavior of individual agents. The prospect of further research is to adapt the model to different types of computing environments, including MEC infrastructures, edge systems, and IoT platforms. Special attention is planned to be paid to the development of new objective functions that would take into account not only the stability and speed of convergence, but also energy consumption, network bandwidth, and quality of service (QoS).
  • Ескіз
    Документ
    Task optimisation in multiprocessor embedded systems
    (Хмельницький національний університет, 2025) Martiniuk, Dmytro; Lyhun, Oleksii; Drozd , Andriy; Besedovskyi, Oleksii
    The relevance of this work lies in the fact that the existing task distribution in multiprocessor embedded systems plays a key role in the development of devices used in various industries. Despite the progress made, there are still many research challenges that require in-depth analysis and implementation of effective solutions. One of the main challenges is to ensure the reliability of embedded systems, especially in environments where safety is critical. Although the functionality of such systems is usually defined at the design stage, ensuring their stable operation in real time remains a challenge. It is necessary not only to guarantee the correctness of calculations, but also to adhere to time constraints, which requires new approaches to managing the resources of multiprocessor systems. Another important problem is the need to meet stringent real-time requirements. This is a characteristic feature of embedded systems, which differ from general-purpose systems that have more flexibility in functionality but do not guarantee such predictability and reliability. Therefore, optimization of task scheduling that takes into account the specifics of embedded systems requires further research. It is also important to take into account the variety of embedded systems, which are divided into control systems and streaming systems that have different data processing requirements. Control systems must respond quickly to environmental events while minimizing delays, while streaming systems process continuous data streams, requiring high throughput and efficiency. The development of universal solutions that can optimize the performance of both types of systems is an urgent task for scientists and engineers. Therefore, task optimization in multiprocessor embedded systems has significant potential for development and is relevant for reliability, real-time guarantees, and efficient resource management, which will contribute to the creation of more secure and productive systems. In this paper, we develop a method for optimizing task execution using replication in a multiprocessor system, which allows to effectively minimize the total execution time, ensure load balance, and minimize communication delays. The peculiarity of the method is the implementation of task migration according to replication using the optimization objective function. An experiment with the system demonstrated that the chosen optimization method effectively balances the load, but additional objective functions are needed to optimize energy consumption. The simulation results show that an increase in the number of processors leads to a decrease in the maximum load and the number of migrations, an increase in the number of tasks increases the system load and the number of migrations at the initial stages, and the migration mechanism effectively balances the load, especially at the initial stages of execution. The areas of further research are the detailing of embedded devices and their classification. For each class of embedded devices, it will be necessary to adapt the algorithms and method of task optimization, as well as to develop the target optimization function.