Computational nanotechnology

This study explores methods for automatically segmenting railway stations using machine learning models by analyzing cargo handling patterns. Automated segmentation techniques help identify key performance indicators and decision-making patterns useful for effective management strategies. Data preprocessing and clustering algorithms are employed with optimized machine learning models. Neural network architectures and deep learning methodologies are developed to classify stations dynamically during operational load monitoring. Practical recommendations are given for designing modular systems capable of intelligently distributing stations, clients, and other entities across targeted segments while accounting for data non-linearities and specific characteristics.

Task. Image blur is one of the most common defects in photography. In the absence of information about the optical system at the time of shooting, it is impossible to accurately determine the blur model and its parameters. For small images, neural network approaches or the use of mathematical methods based on the Radon, Hough transform and the cepstral method are possible. However, in the case of large images, these methods are not applicable due to the high computational complexity, so there is a need to develop a new method for processing such images.

Model. A new method for estimating linear blur parameters on uniformly distorted images is proposed, based on comparing two-dimensional histograms of gradients with pre-calculated histograms obtained from a reference image with various simulated blur parameters. Mostly this method is statistical, using classical mathematical methods of image processing.

Results. The proposed method shows qualitative results for any linear blur parameters. At the same time, the closer the reference image is to the original one in the gradient histogram, the more accurate the result is, up to zero error. If we take as a reference image not the closest one, but from the same class of images, the error as a result will be no more than half a pixel in each of the directions of the blur.

Practical significance. The proposed method is applicable for processing high-resolution images with linear blur. The developed algorithm is applicable, for example, for processing satellite images.

Value. The proposed method has a great advantage over known mathematical methods and neural network methods for determining lubrication parameters due to its high accuracy with low computational complexity. Its application will bring significant benefits in real-time image processing.

The main task in building complex predictive models is to prevent overfitting, especially in small samples, which can lead to incorrect management conclusions. This is due to the high dynamics of the market, limited resources and the nature of information flows (incompleteness, noise) in small and medium-sized enterprises (SMEs), which require the creation of stable and accurate intelligent tools to support management decisions. To solve this problem, data mining uses regularization extensively, which, in fact, imposes restrictions on the complexity of the model. The combination of Bayesian inference and regularization principles forms a regularizing Bayesian approach that creates stable and generalizing models. The role of Bayesian regularization is to select weighting factors for forecasting accuracy, and when generalized, this approach becomes the basis for creating digital platforms for managing complex systems in Industry 4.0 and high-tech SMEs. Thus, the relevance of the research in this article is due to the need to develop an integrated methodological framework that combines the principles of system analysis with a regularizing Bayesian approach. The research aims to contribute to improving the quality of management decisions in SMEs through the introduction of intelligent, statistically sound, and data incompleteness–resistant models that have been integrated into the new adaptive hierarchical regularizing Bayesian model of decision-making. AIRBM successfully implements the principle of system analysis, linking external system uncertainty with the internal configuration of the model, which provides a reliable and stable basis for intelligent management of SMEs in the digital economy.

The increasing volume and heterogeneity of data in the modern digital environment introduce additional challenges to the development of systems for information analysis and security threat detection. Conventional processing approaches lack the flexibility required to adapt to evolving conditions, growing system requirements, and the need for integrated analysis across heterogeneous data sources. This creates a demand for a multimodal data processing constructor capable of dynamically assembling analytical workflows and orchestrating the operation of diverse processing modules. To ensure the correctness and predictability of such a system, it is necessary to formalize its internal processes using graph-based models, which provide a structured representation of computational pipelines, data-flow dependencies and operation-execution rules.

The purpose of the study: to develop an architecture for a multimodal data-processing constructor based on the application of graph models. The proposed architecture is intended to ensure consistent processing of multiple data modalities and to manage the execution sequence of operations within complex analytical workflows.

The research methodology is a combined formalization based on integrating a directed acyclic graph, a colored Petri net, and a state transition graph into a unified formal model that describes the structure, data flows, and operational logic of the multimodal processing constructor. Using the constructed model, an analytical workflow is simulated to evaluate the consistency of processes and the correctness of modality interactions.

Research results: formation of a formal architecture for a multimodal data-processing constructor based on graph models, representing the structure of analytical workflows, data flows, and component behavior. The conducted simulation demonstrated that the system ensures stable modality interaction and correct execution of operations.

Regular construction supervision over the execution of construction and installation work is necessary to ensure that construction proceeds at the established quality level. Of particular importance in this control is establishing precise correspondence between the geometry of structures being erected and their structural elements with the project design. The quality of such control depends on the experience of the expert geodesist. However, even an experienced expert performs many redundant measurements, which complicates and increases the processing time of the received information. For this reason, an autonomous robotic platform (ARP) based on a quadruped walking robot with an installed 3D laser scanner, navigation system, video camera, integration unit, and protective frame was utilized. The article examines the determination of movement trajectory under conditions of high congestion on construction access roads, stopping locations for 3D scanning, and the speed of ARP movement.

The article discusses the issue of improving management processes in organizational systems through the introduction of a regularizing Bayesian approach (RBP). The relevance of the research is due to the need to increase the stability of decisions made in conditions of high uncertainty and dynamic changes in the characteristics of information channels. A mathematical model has been developed that integrates a priori expert knowledge and current monitoring data through the Bayesian inference apparatus using Tikhonov regularizing functionals. The article describes an algorithm for the functioning of the system, including the formation of surrogate models based on Gaussian processes and optimization through the acquisition function. The results of the work were a synthesized intelligent control architecture that minimizes the entropy of the information system and meets the criteria of sustainability. Numerical experiments in twelve scenarios have confirmed the advantage of the proposed method over classical algorithms. The average performance increase was 17.4%, with a significant decrease in variance (to 0.045) and a stability coefficient of 0.94. The advantages of the approach in terms of working with small amounts of data and interpretability of results are formulated, as well as the limitations associated with computing capacity and expert dependence.

There is a significant increase in the number of patients with upper limb dysfunction caused by the social aging of the population. Using a rehabilitation robot for training is sure to become a new trend. In this paper, the object of research is the movement speed of 4-degree-of-freedom upper limb rehabilitation robot. This paper presented a segmentation trajectory planning method for upper limb rehabilitation robots based on quintic polynomials. The algorithm of particle swarm optimization, which solved the problem of how to optimize the motion trajectory of the upper limb rehabilitation robot to ensure the security of rehabilitation training, is presented and successfully verified. In order to give the best training effect to patients, doctors can adjust the working speed of the rehabilitation robot according to the rehabilitation status of the patients’ arm to meet the needs of patients for rehabilitation training in different periods of rehabilitation. The simulation results from Matlab software confirm the feasibility of this method.

The annotating metagraph model is a powerful tool for describing complex systems with hierarchy and emergent properties, yet its syntactic representation poses a non-trivial challenge.

The purpose of this work is to investigate the problem of representing this model using popular graph description languages: DOT, GraphML, and JSON Graph Format. A comparative analysis of their syntactic capabilities is conducted by encoding metagraphs that contain nested and overlapping metavertices.

The research concludes that these languages face fundamental limitations for metagraph description. Their syntax, based on a strict tree-like structure, prevents the accurate representation of key metagraph features, such as an element’s membership in multiple, overlapping metavertices. To address this issue, the paper proposes a new, specialized format called MetagraphYAML. Its id-based referencing system fully supports all capabilities of the annotating metagraph model, thereby overcoming the limitations of existing approaches.

This paper examines a method for establishing secure communication channels on online platforms using post-quantum cryptography. It is shown that quantum attacks on communication channels can compromise the security of banking data, critical information infrastructure, and state secrets. Therefore, when establishing a communication channel, the protection of registration data exchange should be built using post-quantum cryptography. The paper proposes using the CRYSTALS cryptosystem in combination with Shamir’s secret sharing scheme during the registration phase. The latter is used to securely distribute the “seed,” a unique random sequence from which cryptographic keys are generated. Dividing the “seed” among multiple nodes eliminates a single point of compromise and reduces the risk of internal threats, including abuse by administrators. The proposed method allows users to securely restore access on additional devices. The platform for building communication channels is based on a federated architecture that ensures secure messaging between branches while maintaining core data within trusted networks. Client applications verify the integrity of server responses and the application itself, preventing man-in-the-middle attacks. The growing number of online communication platforms requires the implementation of reliable mechanisms for secure user registration, especially in environments handling sensitive information. The proposed solution provides a quantum-resistant and fault-tolerant method for establishing secure communication channels, guaranteeing the confidentiality, verifiability, and recoverability of user registrations across devices.

Building on the previously developed method for adaptive threat detection and a software prototype of a digital twin for automated power grid control systems, this paper presents a comparative analysis of three strategies for maintaining the relevance of information security threat detection models for critical information infrastructure (CII) objects. The approach is validated using the example of an automated control system for an intelligent power grid. The examined strategies include: a static strategy (without model updates), a strategy of retraining on real data, and an adaptation strategy using a digital twin, where model updates are based on synthetic data generated in a virtual environment. Experimental evaluation was conducted on a simulation platform that reproduces telemetry and typical operating modes, as well as models of cyberattacks: imitation of normal changes, pulse attacks, and combined attacks. The results demonstrate that the static model is incapable of detecting new types of threats, while both adaptive strategies provide high detection recall. The strategy employing a digital twin achieves the highest recall with a comparable level of false positives, while simultaneously minimizing the use of data from the real object and reducing operational risks.

The article considers the task of increasing the flexibility and manageability of the processes of building data storefronts in intelligent transport systems. The relevance of the research is due to the need for rapid changes in data processing algorithms in the context of dynamically changing requirements and a large volume of incoming information from transport infrastructure and telematics sources. A mechanism for managing algorithms for calculating data marts is proposed, providing centralized storage, validation, and dynamic application of computational rules without modifying the main code of ETL processes. The architecture of the software solution has been developed, including a user interface, a server application, and a repository of rules in a database, as well as the integration of the mechanism into the processes of building data marts. A mathematical model of the application of computational rules is presented, formalizing the process of selecting and composing data processing functions. Functional testing of the developed mechanism was carried out, which confirmed the correctness of its operation and the possibility of using it in various operating modes, including the pilot implementation of algorithms. The practical significance of the work lies in the possibility of using the proposed mechanism in intelligent transport information systems to reduce the time needed to implement algorithm changes and increase the reliability of data processing processes.

The article offers a reproducible methodology for selecting the architecture of intelligent information systems (AIS) for additional professional education (APE). The classification of architectures (centralized, distributed/microservice, cloud, multi-agent) was performed, six evaluation criteria (adaptability, scalability, interoperability, reliability, data security, complexity of implementation) were formalized, and an integrated metric was developed that takes into account external constraints (technical, legal, financial) and the phase aspect of the application (entry, mass training, individual support). The methodology is based on an analytical review of publications, comparative analysis and case analysis of typical platforms (Moodle, Open edX, Stepik), as well as the calculation of a summary indicator of the suitability of architectures. The scientific novelty consists in the integration of architectural, technical, pedagogical and technological requirements into a single model with explicit consideration of limitations and phases of use; the practical significance lies in the possibility of adjusting weights and the applicability of the model for decision-making in vocational training organizations. The work forms a methodological basis and serves as a starting point for subsequent scientific research and empirical validation of the proposed model.

Quantum networks represent a promising and rapidly developing field at the intersection of quantum physics and information technology, intended to serve as the architectural foundation for the quantum Internet. The paper examines the fundamental principles of quantum network construction, including the use of qubits and photonic systems for transmitting quantum information over fiber-optic channels. A comparative analysis of classical (LAN, WAN, SAN) and quantum (QLAN, QWAN, QSAN) computing network models is carried out. Various types of quantum sensor networks and examples of their scientific applications are described, including quantum magnetometer networks, entangled atomic clock networks, and phase-sensitive networks. The concepts of blind quantum computing, which ensures the confidentiality of delegated computations, and quantum key distribution protocols as a basis for cryptographic protection of communication channels are considered. Key technological barriers related to the no-cloning theorem and the need for developing quantum repeaters are identified. A computational experiment simulating the BB84 protocol is reported, yielding quantitative error rate estimates for varying channel noise and eavesdropping levels.

This article analyzes new renewable energy sources by energy type and quantitative parameters. Options for improving battery cost and safety are discussed. Optimal service life for sodium-sulfur batteries are determined. The problems determining battery life and the possibilities for extending their service life are discussed. The disadvantages of sodium-sulfur batteries, as well as their distinctive and attractive features, are highlighted. It is concluded that sodium-sulfur batteries are inexpensive, high-capacity, non-explosive, and safe.

This study presents the development and spectral characterization of an asymmetric three-layer polyethylene (PE) composite film integrated with ceramic granules for solar drying applications. The primary objective was to investigate the influence of ceramic layer orientation on the optical transmittance, reflectance, and absorption properties within the UV-Vis-NIR range 300–1100 nm. Experimental results revealed that the composite film exhibits significant optical asymmetry compared to pristine PE. When correctly oriented (ceramic layer facing the product), the film demonstrates superior UV-blocking (reduction from 82.4 to 12.6%) and enhanced NIR absorption, facilitating the generation of pulsed infrared radiation. Conversely, reverse orientation significantly diminishes these effects due to increased surface reflection and thermal leakage. The findings confirm that the 3-layer asymmetric structure acts as a ’photonic valve,’ optimizing solar energy harvesting and thermal conversion. This research provides a technical foundation for utilizing directionally-active composite films to improve the efficiency and quality of solar-based food dehydration processes.