Information, Computing and Intelligent systems

Neural network model for autonomous navigation of a water drone

2024-11-20T11:17:01+02:00

Water drones have significant potential for use in environmental monitoring, search and rescue operations, and marine infrastructure inspection, but the specific conditions of the water environment make it difficult to implement stable autonomous navigation.

The object of research presented in this paper is the machine learning process for autonomous navigation of a water drone model in a simulated water environment. The purpose of the study is to implement a neural network model for autonomous navigation of a water drone using a reinforcement learning method that provides improved obstacle avoidance and adaptation to water currents.

To achieve this purpose, a new neural network model for autonomous drone navigation in the water environment based on the reinforcement learning method is proposed, which differs from the existing ones in that it uses an improved drone control algorithm that takes into account the speed and direction of the water current, which makes it possible to stabilize the process of generating neural network coefficients.

To ensure an effective learning process and optimization of the model, a simulation training environment was developed using the USVSim simulator, which contains various factors that interfere with the drone's movement, such as water current and the presence of other objects. The water drone, acting as an agent, gradually learns to choose the most effective actions to maximize positive rewards through trial and error, interacting with the environment and adapting to changing conditions. This process takes place through the use of a Deep Q-Network: the drone provides the value of its current state to a deep neural network; the neural network processes the data, predicts the value of the most effective action, and gives it to the agent. The current state of the drone is information in the form of a set of sensor readings measuring the distance to the nearest obstacles, drone’s heading and current distance to goal. The value of the effective action received from the neural network is converted into a command for the rudder that the drone can understand. The value of the drone's thruster power is calculated by separate formulas using trigonometric functions.

The results of the study showed that the use of the proposed model allows the drone to make decisions in a dynamic water environment when rapid adaptation to changes is required. The model successfully adapted its strategy based on feedback from the environment, so it can be concluded that the implemented model shows significant potential for further research and applications in the field of autonomous water drones, especially in changing and unpredictable environments.

Software for collecting and analyzing metrics in highly loaded applications based on the Prometheus monitoring system

2024-11-28T18:22:10+02:00

This paper emphasizes the importance of collecting metrics during application operation for early detection of potential problems. The undisputed leader in this area is the Prometheus monitoring system, which, combined with Grafana – a platform for visualizing collected data in numerous graphs – becomes an indispensable tool for programmers and site reliability engineers. However, the average value of a certain metric is often unrepresentative, because it does not reflect a comprehensive picture. Instead, collecting metrics in terms of various quantiles over a long period is useful to identify even single instabilities. Still, the use of standard tools in the Python ecosystem may require a lot of server resources and long preliminary analysis, which can be quite costly for businesses from a financial point of view. That is why the development of a new approach for collecting and analyzing metrics in highly loaded applications based on the Prometheus monitoring system is relevant.

The research aims to improve the efficiency of storing metrics across different quantiles, which will create additional opportunities for further analysis.

A review of existing approaches for calculating quantile values on large data sets was conducted. Their comparative characteristics in terms of speed and memory usage were also presented. The chosen method was adapted for use with the real-time data stream and implemented as a Python extension for the official Prometheus library. It opens up opportunities for comprehensive monitoring of highly loaded systems in terms of both server resource usage and the quantity and quality of collected useful data. This solution can be easily implemented on large projects requiring continuous tracking of various metrics to ensure stable and uninterrupted service operation.

Models and methods for forming service packages for solving of the problem of designing services in information systems of providers

2024-11-29T16:01:58+02:00

Today, in the telecommunications industry, service is one of the fundamental concepts. Building a service architecture is a key stage in the service life cycle. Information systems of telecommunications providers are designed, implemented and supported by IT companies based on the End-to-End model. This requires the IT company to solve a number of complex problems. In these conditions, building a service architecture, implementing and providing a service are divided into subproblems. Solutions to subproblems must be integrated to determine the coordinated activities of both the IT company and the provider. In this case, it is necessary to take into account the goals of IT companies, providers and their customers in such a way that it is beneficial to all parties. One of such subproblems is the formation of service packages that the IT company offers to providers. The article proposes formal models for the subproblem of forming service packages that allow taking into account the interests of the IT company and providers. These are multi-criteria nonlinear mathematical programming models. To solve the subproblem of forming service packages, a two-stage algorithm and a modified version of the guided genetic algorithm are proposed. The use of these methods allows us to take into account the interests of the IT company and providers. Also, such important factors that affect the formation of packages as the base price of the service, service dependency, discount system, resource and other constraints of the IT company and providers are taken into account.

The two-stage algorithm at the first stage uses classical algorithms for solving the knapsack problem, and at the second stage implements a compromise scheme to improve the solution. The second of the proposed methods uses three types of tools in combination. The first tool controls the convergence of the genetic algorithm. The second tool determines the choice of the best solutions taking into account the features of the multi-criteria problem. The third tool allows to obtain the best solutions to the optimization problem with the simultaneous choice of a discount strategy. Experimental studies have confirmed the effectiveness of the proposed methods. Their ability to form the basis of the technology of forming service packages as a component of the platform for supporting the life cycle of services is also confirmed.

Method of horizontal pod scaling in Kubernetes to omit overregulation

2024-11-21T21:35:43+02:00

This paper describes the method of omitting the over-regulation effect that occurs under certain conditions by horizontal pod autoscaling microservices in the container application orchestration system Kubernetes. The effect was initially observed only for long-term HTTP WebSocket sessions, where it led to excessive use of computing resources, which reduced the efficiency of IT infrastructure management, and caused service failure. It was found that the overregulation effect is reproduced not only for connections with long-term HTTP sessions, such as HTTP WebSocket, but also for shorter-term REST HTTP sessions in case of increased delay in the metric collection cycle used for horizontal pod autoscaling. It is assumed that this effect happens due to the approach of implementing horizontal scaling controllers similar to the principles of proportional regulators in systems with negative feedback from the theory of automation and control. It is proposed to extend one of the methods used for optimizing the proportional controller to the problem consisting of reducing the time delay between scaling metrics collecting and upscale applied by the controller in Kubernetes. The applied method demonstrated its effectiveness, therefore, within the same methodology, an experiment was conducted on using the proportional-integral-differential controller for automatic horizontal scaling of pods. The results obtained showed why the proportional-integral-differential controller is not widespread among the overviewed Kubernetes solutions for horizontal automatic scaling. An assumption was made about the limitations of studying the downscaling process in Kubernetes due to the need to consider the quality of service when stopping pods and the need to collect indicator metrics using quality-of-service object management tools such as ISTIO.

Effectiveness of Hybrid Quantum-Classical and Quanvolutional Neural Networks for image classification

2024-12-23T16:30:30+02:00

The article focuses on studying the effectiveness of two different Hybrid Neural Networks (HNNs) architectures for solving real-world image classification problems. The first approach investigated in the research is a hybridization technique that allows creation of HNN based on a classical neural network by replacing a number of hidden layers of the neural network with a variational quantum circuit, which allows to reduce the complexity of the classical part of the neural network and move part of computations to a quantum device. The second approach is a hybridization technique based on utilizing quanvolutional operations for image processing as the first quantum convolutional layer of the hybrid neural network, thus building a Quanvolutional Neural Network (QNN). QNN leverages quantum phenomena to facilitate feature extraction, enabling the model to achieve higher accuracy metrics than its classical counterpart.

The effectiveness of both architectures was tested on several image classification problems. The first one is a classical image classification problem of CIFAR10 images classification, widely used as a benchmark for various imagery-related tasks. Another problem used for the effectiveness study is the problem of geospatial data analysis. The second problem represents a real-world use case where quantum computing utilization can be very fruitful in the future. For studying the effectiveness, several models were assembled: HNN with a quantum device that replaces one of the hidden layers of the neural network, QNN based on quanvolutional operation and utilizes VGG-16 architecture as a classical part of the model, and also an unmodified VGG-16 was used as a reference model. Experiments were conducted to measure the models' key efficiency metrics: maximal accuracy, complexity of a quantum part of the model and complexity of a classical part of the model.

The results of the research indicated the feasibility of both approaches for solving both proposed image classification problems. Results were analyzed to outline the advantages and disadvantages of every approach in terms of selected key metrics. Experiments showed that QNN architectures proved to be a feasible and effective solution for critical practical tasks requiring higher levels of model prediction accuracy and, simultaneously, can tolerate higher processing time and significantly increased costs due to a high number of quantum operations required. Also, the results of the experiments indicated that HNN architectures proved to be a feasible solution for time-critical practical tasks that require higher processing speed and can tolerate slightly decreased accuracy of model predictions.

Depth-Width type criteria approbation for tree shape control for the Monte Carlo Tree search method

2024-12-09T21:27:31+02:00

This paper is devoted to the scientific problem of improvements of the Monte Carlo Tree Search (MCTS) method. The object of research is the process of performing a tree search using the MCTS. The subject of research is the MCST improvement technique with control of the search tree shape by usage of the previously proposed be the authors DWC (Depth/Width Criterion) and WDC (Width/Depth Criterion) criteria. This technique was named Monte Carlo Tree Search with Tree Shape Control (MCTS-TSC). The research methods are based on the theory of data structures and analysis methods.

The aim of the study is to conduct extended study of the previously proposed MCTS-TSC technique for improvement of the MCTS method. In particular, the aim is to approve that the DWC and WDC tree shape control criteria ensure the better move selection and increasing player strength compared to the standard Monte Carlo Tree Search with Upper Confidence bounds applied to Trees (MCTS-UCT) technique.

To achieve the aim, the following tasks were set: to conduct a set of experiments according to the developed approbation methodology to approve that the WDC criterion of the MCTS-TSC technique is able to improve the MCTS method; to conduct a set of experiments according to the developed approbation methodology to approve that the DWC criterion of the MCTS-TSC technique is able to improve the MCTS method.

Both WDC and DWC criteria of the MCTS-TSC technique were tested on a series of games of Connect Four between a player, which used the MCTS-TSC technique, and a player which used the MCTS-UCT technique. Different parameters for tuning the formulas of the WDC and DWC criteria of the MCTS-TSC technique were used in the experiments.

The paper describes the methodology of the approbation of the MCTS-TSC technique with usage of the WDC and DWC criteria compared to the MCTS-UCT technique and conducts comparative analysis of the results of the experiments. The MCTS-TSC player won from 30% to 70% more games than the MCTS-UCT player for some search tree shapes, when WDC criterion was used, and from 19% to 52% more games, when DWC criterion was used. So, ability of the proposed
MCTS-TSC technique to improve the MCTS method was approved for both criteria, WDC and DWC.

Key words: depth-width type criteria, Monte Carlo tree search method, MCTS, MCTS-UCT,
MCTS-TSC, search tree shape control.

Dynamic mathematical model for resource management and scheduling in cloud computing environments

2024-12-01T23:29:14+02:00

The object of the research is resource management and scheduling in Kubernetes clusters, in particular, data centers. It was determined that in many publications dedicated to optimization models of scheduling for Kubernetes, mathematical models either do not include constraints at all, or only have the constraints determined on the high level only. The purpose of the research is the creation of a dynamic low-level mathematical optimization model for resource management and scheduling in cloud computing environments that utilize Kubernetes. Examples of such environments include the data centers where the customers can rent both dedicated servers and resources of shared hosting servers that are allocated on demand. The suggested model was created using the principles of creation of mathematical models of discrete (combinatorial) optimization, and was given the name “dynamic” because it takes the time parameter into account.

The model receives data about individual servers in the cluster and individual pods that should be launched as an input. The model aims to regulate not only individual assignments of pods to nodes, but also turning on and off the servers. The model has objectives of: minimization of the average number of shared hosting servers running; maximization of the average resource utilization coefficient on such servers; minimization of the number of occasions when the servers are turned on and off; minimization of resource utilization by the pods that are running on shared hosting servers but created by the customers renting the dedicated servers. The model considers resource constraints, among other limitations.

The algorithm for selecting publications on a given topic considering keyword priorities

2024-12-01T01:51:39+02:00

The article investigates the problems that exist in existing search engines for scientific publications. The search algorithms used in various search engines for scientific publications are described. The aim of the article is to develop a method for selecting publications on a given topic based on assessing the relevance of keyword sets. A review of the literature that was analyzed during the research is presented. Among the publications studied were materials related to the theory of set similarity, namely the use of the Jacquard coefficient and editing distance. A measure for determining the similarity of keyword sets is presented, which is based on the Jacquard coefficient taking into account the weighting coefficients of keywords. An algorithm is presented that can be used to determine the degree of similarity of publications to a user's search query based on keyword sets with weighting coefficients. The algorithm is based on the measure of similarity presented by us and the editing distance presented by us. The algorithm can be used to rank search results in search engines for scientific publications, as well as to compare the efficiency of different search engines, assess the quality of the results they return. The algorithm can also be used in book and film recommendation systems based on user preferences. The article provides the pseudocode of the algorithm. It is demonstrated on a limited data set how the measure calculated by the algorithm changes depending on the distribution of keyword weights in the user's query and the number of keywords.

The automatic cryptocurrency trading system using a scalping strategy

2024-12-02T14:50:53+02:00

The study focuses on the development and implementation of an automated system for scalping strategies in cryptocurrency markets. Scalping, a high-frequency trading strategy, aims to generate profits from small price fluctuations. The primary goal of the research is to create an automated trading bot that addresses critical issues such as latency, risk management, scalability, and reliability in
real-world market conditions. To achieve this, the following objectives were defined: develop a novel scalping method, implement a software solution to integrate the method into an automated trading system, and evaluate its effectiveness through experimental testing.

The research methodology utilized technical indicators, including the Exponential Moving Average (EMA) and Volume Weighted Average Price (VWAP). Pseudocode was created to illustrate the decision-making process, incorporating key parameters such as smoothing factors, time periods, and thresholds for trade execution. The software architecture consists of modules: Binance exchange integration, data collection and management, strategy analysis, trade execution, and historical data storage. Technologies such as PostgreSQL, Redis, WebSocket, and Python libraries (Pandas, NumPy, TA-Lib) were employed to ensure the robustness and efficiency of the system.

Experiments were conducted using the BTC/USDT trading pair, known for its high liquidity and volatility. The system was tested on hardware featuring an Intel Core i7-10700K processor, 32 GB of RAM, and a 1 Gbps network connection. A comparative analysis between the scalping strategy and a trend-following strategy demonstrated the advantages of scalping in volatile markets. The scalping bot executed 15 trades (13 successful) within two hours, achieving a total profit of 120 USDT.

Performance metrics, including latency (15–50 ms), signal processing time, CPU utilization
(5–55%), and memory usage (120–2100 MB), were measured. The results confirmed the system's modular architecture and its ability to scale linearly with increasing trading volumes.

The findings validate the effectiveness of the proposed method and the reliability of the developed system in real-world conditions. Future research may focus on optimizing algorithms to reduce resource consumption and integrating advanced risk management techniques to enhance performance.

A method and software for license plate recognition

2024-12-01T22:40:12+02:00

В статті представлений метод розпізнавання номерних знаків із використанням сегментації шляхом використання системи детектування YOLO у поєднанні із завдання-орієнтованим підходом до процесу навчання та використанням масивів варіативних даних реального світу.

Розвиток мегаполісів і постійне збільшення кількості транспортних засобів на дорогах призвели до нового рівня вимог до систем безпеки дорожнього руху. Автоматизація, без перебільшення, є найбільш пріоритетним напрямком розвитку цих систем. Лише за допомогою автоматизації системи безпеки дорожнього руху можуть обробляти величезну кількість інформації, що генерується на дорогах щодня. Крім того, автоматизація дозволяє поступово зменшувати участь людини в задачах, які обчислювальні системи можуть виконувати з еквівалентною або більшою точністю. Ці досягнення спрямовані на мінімізацію впливу людського фактору, а також на зниження експлуатаційних витрат. Це особливо важливо для мегаполісів, але також стосується транспортної системи в цілому.

Метою дослідження є розробка методу автоматизованого розпізнавання номерних знаків для підвищення точності систем забезпечення дорожньої безпеки шляхом зниження рівня помилок, мінімізації надмірного використання обчислювальних ресурсів у процесі виявлення та здешевлення таких систем. Об’єктом дослідження є процес розробки автоматизованих програмних систем для забезпечення дорожньої безпеки з інтеграцією функціоналу ідентифікації транспортних засобів.

Для досягнення поставленої мети були визначені такі завдання: розробити метод розпізнавання номерних знаків із застосуванням цілеспрямованого підходу до навчання у поєднанні з системою виявлення YOLO; оцінити вплив попередньої сегментації номерних знаків із використанням спеціально навченої системи YOLO на рівень помилок і часові витрати, а також провести експерименти із застосуванням запропонованого методу навчання на реальних зображеннях із варіативним довкіллям для підтвердження його адекватності.

Порівняльний аналіз використання завдання-орієнтованого методу навчання системи детектування на базі YOLO v5 лише з загальноприйнятим методом оптичного розпізнавання символів (Optical Character Recognition, OCR) підтвердив переваги завданняорієнтованого методу при вирішенні завдання з розпізнавання номерних знаків. Також було досліджено вплив розмиття на результати детектування із використанням OCR методу.

Результати практичних досліджень підтверджують правильність обраних методів для підвищення ефективності розпізнавання номерних знаків.

Dynamic model of currency exchange based on investor behavior

2024-11-30T00:19:53+02:00

In the modern financial environment, cryptocurrencies have gained significant popularity, becoming an important element of the global economy and financial markets. The dynamic development of blockchain technologies and decentralized financial instruments fosters increased interest from both private investors and institutional players. However, the high volatility of cryptocurrencies and the complexity of the mechanisms behind their price formation necessitate a detailed study of these processes.

This paper models cryptocurrency exchange operations, analyzing price formation influenced by buying, selling, and introducing new crypto coins to the market. The system simulates investor behavior with individual parameters: initial balances, risk profiles, and profit-driven trading strategies over a specified period. The model takes into account the psychological aspects of investor behavior, their reaction to changing market conditions, and the impact of external factors such as news and regulatory changes.

Special attention is paid to analyzing the impact of adding additional quantities of coins to the exchange at a reduced price during peak cryptocurrency price values. This creates conditions for activating trading operations, increasing liquidity, and affecting overall market dynamics, particularly volatility and price fluctuation trends. The study shows how such interventions can be used to stabilize the market or stimulate its further growth.

The analysis of the obtained data allows for detailed observation of changes in the cryptocurrency’s value over time, identifying patterns and trends. Using statistical and analytical methods, the impact of different investor strategies on their financial results and the overall market situation was investigated. This enables assessing how investor decisions-timing, trade volume, and market reactions-impact profits and market dynamics.

The research emphasizes the importance of a deep understanding of market mechanisms and trading psychology and can serve as a basis for developing effective trading strategies on cryptocurrency exchanges. The obtained results may be useful for traders, financial analysts, and developers of algorithmic trading systems, contributing to increased efficiency and stability of cryptocurrency markets. Moreover, the findings of the work can be applied to improve regulatory approaches and policies regarding cryptocurrencies.

Scientific article summarization model with unbounded input length

2024-11-06T10:45:07+02:00

In recent years, the exponential growth of scientific literature has made it increasingly difficult for researchers and practitioners to keep up with new discoveries and developments in their fields. Thanks to this, text summarization has become one of the primary tasks of natural language processing. Abstractive summarization of long documents, such as scientific articles, requires large neural networks with high memory and computation requirements. Therefore, it is all the more important to find ways to increase the efficiency of long document summarization models.

The objects of this research are long document summarization transformer models and the Unlimiformer cross-attention modification. The article reviews the basic principles of transformer attention, which constitutes the primary computational expense in transformer models. More efficient self-attention approaches used for long document summarization models are described, such as the global+sliding window attention used by Longformer. The cross-attention mechanism of Unlimiformer, which allows a model to have unbounded input length, is described in detail. The objective of the study is the development and evaluation of a long document summarization model using the Unlimiformer modification. To achieve this goal, a Longformer Decoder-Encoder model pretrained on the arXiv dataset is modified with Unlimiformer cross-attention. This modification can be applied without additional model fine-tuning, avoiding the cost of further training a large sequence length model.

The developed model was evaluated on the arXiv dataset using the ROUGE-1, ROUGE-2 and ROUGE-L metrics. The developed model showed improved results compared to the baseline model, demonstrating the viability of using this approach to improve long document summarization models.

Improving the effectiveness of monolith architecture to microservices migration using existing migration methods

2024-11-27T21:14:34+02:00

The theme of the transition from monolithic architecture to microservice is one of the key challenges of modern software engineering. This transformation allows for greater flexibility, scalability and adaptability of systems, but requires careful planning and consideration of numerous factors that affect the efficiency of migration. This study aims to improve the algorithm for determining the effectiveness of using methods for migrating monolithic systems to microservice architecture. Migration from monolithic architecture to microservice is a complex process involving significant technical and organizational challenges. Since monolithic systems often have a complex structure and relationships between components, the transition to a microservice architecture requires careful planning and selection of effective migration methods. The lack of a unified approach to assessing the effectiveness of different migration patterns makes the transition process difficult and risky.

The article is aimed at improving the algorithm for determining the efficiency of using migration methods from monolithic architecture to microservices. To do this, we compare existing migration patterns, such as the Strangler Fig Pattern, Branch by Abstraction, Parallel Run, Decorating Collaborator and Change Data Capture, according to the criteria: implementation time, test complexity, error risk, performance degradation and efficiency. The study uses methods of comparative analysis and quantitative evaluation of the effectiveness of migration patterns. For this, criteria are applied to assess the implementation time, testing complexity, possible risks, as well as the impact on system performance. In addition, scenarios are analyzed in which each template is most effective, which allows you to determine the optimal approaches to migration depending on the specifics of the project.

The obtained results allow not only a deeper understanding of the advantages and disadvantages of different approaches to migration, but also to form recommendations for choosing the optimal pattern, depending on the specifics of the system and business needs. The scientific novelty of the study is to create an algorithm that integrates these criteria to increase the efficiency of migration processes. The results of the work can be useful for software engineers, architects and managers planning the transition to microservice architecture, providing a structured methodology for evaluating and selecting migration methods.

CI/CD integration tools for automated code deployment and verification for training purposes

2024-12-23T22:24:29+02:00

The article is devoted to the study and application of modern tools for Continuous Integration and Continuous Deployment (CI/CD) in the educational field. Automating the processes of software deployment and testing is a relevant task for both improving the educational process and developing DevOps skills among students. Significant attention is given to studying the core principles of CI/CD, including automated testing, code quality monitoring, and integration with source code repositories.

Popular CI/CD platform such as Jenkins is utilized to automate the educational process and train students. This tool enables the creation and deployment of applications using Docker technologies, which allow real-world scenarios to be modeled. A significant emphasis is placed on the scalability and adaptability of solutions, which enhance the efficiency of resource usage.

A methodology for implementing CI/CD into an educational course is proposed, including integration with project management platforms and version control systems such as Git, with Gitea as an example. The main stages include setting up automated builds, testing, and deployment, which enable students to practice the principles of continuous integration and delivery. From the perspective of improving the efficiency of the educational process, the proposed methodology allows for the automation of assignment verification. The problems of Gitea and Jenkins integration are considered. A way for integrating these tools through locally installed Jenkins and Gitea with private code repositories has been proposed. Recommendations are provided for organizing the educational process through practical and laboratory work focused on real-world scenarios of software deployment and test automation.

The results of the study confirm the effective use of CI/CD tools for educational purposes, ensuring the development of competencies required for working in modern IT teams. The use of CI/CD increases awareness of cybersecurity and optimizes DevOps processes.