System research and information technologies

Determination of the generalized optimality criteria for selecting civilian shelter facilities from attacks by ballistic (cruise) missiles and kamikaze drones in urbanized areas

Vadim Yakovenko — 2024-09-28

The object of the study is the planning of the selection of civilian shelter from attacks by ballistic (cruise) missiles and kamikaze drones in urbanized areas. A generalized model for assessing the choice of civilian shelter facilities has been developed by applying linear forms of factor linkage in combination with a generalized optimality criterion in the form of a linear combination of local criteria. The multivariate regression analysis method was chosen to study the correlation between the generalized criterion and the observed feature. A generalized criterion for the optimal choice of civilian shelter facilities from attacks by ballistic (cruise) missiles and kamikaze drones in urbanized areas is calculated in the form of regression coefficients. The criterion can facilitate a simplified determination of the generalized indicator of a linear model for planning the protection of civilians in cities outside the area of hostilities. The initial data is a set of physical (technical) states of shelters with a list of values and features sufficient to assess their resistance to high dynamic loads.

Data scrambler knight tour algorithm

Vadim Romanuke — 2024-09-28

Nowadays, data scrambling remains a vital technique to protect sensitive information by shuffling it in a way that makes it difficult to decipher or reverse-engineer while still maintaining its usability for legitimate purposes. As manipulating the usability of the scrambled data remains a challenge on the background of risking losing data and getting them re-identified by attackers, scrambling and descrambling should be accomplished faster by not increasing data loss and re-identification risks. A scrambling algorithm must have a linear time complexity, still shuffling the data to minimize the risks further. A promising approach is based on the knight open tour problem, whose solutions appear like a random series of knight positions. Hence, a knight open tour algorithm is formalized, by which the knight seems to move chaotically across the chessboard. The formalization is presented as an indented pseudocode to implement it efficiently, whichever programming language is used. The output is a square matrix representing the knight open tour. Based on the knight tour matrix, data scrambler and descrambler algorithms are presented in the same manner. The algorithms have a linear time complexity. The knight-tour scrambling has a sufficiently low guess probability if an appropriate depth of scrambling is used, where the data is re-scrambled repetitively. The scrambling depth is determined by repetitive application of the chessboard matrix, whose size usually increases as the scrambling is deepened. Compared to the pseudorandom shuffling of the data along with storing the shuffled indices, the knight-tour descrambling key is stored and sent far simpler yet ensures proper data security.

Generalized scenarios of transition to chaos in ideal dynamic systems

Oleksii Horchakov — 2024-09-28

The implementation of a new scenario of transition to chaos in the classical Lorenz system has been discovered. Signs of the presence of an implementation of the generalized intermittency scenario for dynamic systems are described. Phase-parametric characteristics, Lyapunov characteristic exponents, distributions of invariant measures, and Poincaré sections are constructed and analyzed in detail, which confirm the implementation of the generalized intermittency scenario in an ideal Lorenz system.

Improving the accuracy of neural network exchange rate forecasting using evolutionary modeling methods

Serhii Fedin — 2024-09-28

A set of models of feedforward neural networks is created to obtain operational forecasts of the time series of the hryvnia/dollar exchange rate. It is shown that using an evolutionary algorithm for the total search of basic characteristics and a genetic algorithm for searching the values of the matrix of neural network weight coefficients allows optimizing the configuration and selecting the best neural network models according to various criteria of their training and testing quality. Based on the verification of forecasting results, it is established that the use of neural network models selected by the evolutionary modelling method increases the accuracy of forecasting the hryvnia/dollar exchange rate compared to neural network models created without the use of a genetic algorithm. The accuracy of the forecasting results is confirmed by the method of inverse verification using data from different retrospective periods of the time series using the criterion of the average absolute percentage error of the forecast.

The role of generative artificial intelligence (GAI) in scientific research

Anatolii Petrenko — 2024-09-28

The emergence and growing capabilities of Generative Artificial Intelligence (GAI) are profoundly transforming scientific research. Although AI extends human intelligence by automating certain tasks, it complements rather than replaces human creativity. This article discusses the implications of AI for the scientific process, including ethical considerations and the need for a balanced approach that combines the strengths of human and artificial intelligence in the process of discovering knowledge and solving complex problems. The discussion extends to the need for universities to adapt their curricula to prepare future researchers for the AI era, emphasizing scenario-based thinking and uncertainty management as important skills for the future.

Reducing risk for assistive reinforcement learning policies with diffusion models

Andrii Tytarenko — 2024-09-28

Care-giving and assistive robotics, driven by advancements in AI, offer promising solutions to meet the growing demand for care, particularly in the context of increasing numbers of individuals requiring assistance. It creates a pressing need for efficient and safe assistive devices, particularly in light of heightened demand due to war-related injuries. While cost has been a barrier to accessibility, technological progress can democratize these solutions. Safety remains a paramount concern, especially given the intricate interactions between assistive robots and humans. This study explores the application of reinforcement learning (RL) and imitation learning in improving policy design for assistive robots. The proposed approach makes the risky policies safer without additional environmental interactions. The enhancement of the conventional RL approaches in tasks related to assistive robotics is demonstrated through experimentation using simulated environments.

Comparison of methods for interpolation and extrapolation of boundary trajectories of short-focus electron beams using root-polynomial functions

Igor Melnyk — 2024-09-28

The article considers and discusses the comparison of interpolation and extrapolation methods of estimation of the boundary trajectory of electron beams propagated in ionized gas. All estimations have been computed using root-polynomial functions to numerically solve a differential-algebraic system of equations that describe the boundary trajectory of the electron beam. By providing analysis, it is shown and proven that in the case of solving a self-connected interpolation-extrapolation task, the average error of the beam radius estimation is generally smaller. This approach was especially effective in estimating the focal beam radius. An algorithm for solving self-connected interpolation-extrapolation tasks is given, and its efficiency is explained. Corresponding graphic dependencies are also given and analyzed.

Identification of nonlinear systems with periodic external actions (Part I)

Viktor Gorodetskyi — 2024-09-28

The problem of identifying nonlinear systems with periodic external actions is considered in the article. The number of such actions in the system is not limited, and these actions can be either additive or multiplicative. We use a time series of observed system variables to calculate unknown equation coefficients. The proven theorem allows us to separate the unknown coefficients of the system into variables and constants. The proposed computational procedure allows us to avoid possible errors caused by the discrete nature of observable time series. Identification of zero coefficients is carried out in two ways, eliminating erroneous zeroing of the terms of the equations. The method is illustrated with a numerical example of identifying a chaotic system with periodic external actions.

Fairness of 2D corotational beam spline as compared with geometrically nonlinear elastic beam

Igor Orynyak — 2024-09-28

The goal of this paper is to further investigate the properties and advantages of corotational beam spline, CBS, as suggested recently. Emphasis is placed on the relatively simple task of drawing the spline between two endpoints with prescribed tangents. In the capacity of “goodness” of spline, the well-known notion of “fairness” is chosen, which presents itself as the integral from the squared curvature of spline over its length and originates from the elastic beam theory as the minimum of energy of deformation. The comparison is performed with possible variants of the cubic Bezier curve, BC, and geometrically nonlinear beam, GNB, with varying lengths. It was shown that CBS was much more effective than BC, where any attempt to provide better fairness of BC by varying the distances from endpoints to two intermediate points generally leads to lower fairness results than CBS. On the other hand, GNB, or in other words, the elastica curve, can give slightly better values of fairness for optimal lengths of the inserted beam. It can be explained by the more sophisticated scientific background of GNB, which employs 6 degrees of freedom in each section, compared with CBS, which operates only by 4 DoF.