Informacionnye Tehnologii

The article proposes a time series forecasting model designed for unstable and partially observable environments. Unlike traditional approaches, the developed FELAR architecture combines local agent learning with reward adjustment based on collective characteristics (trust, reputation, influence), alongside global evolutionary adaptation of strategies. The proposed model operates in a distributed multi-agent environment, enabling both local adaptive behavior and global strategy evolution. A set of experiments on publicly available time series datasets (urban traffic, transformer temperatures, electricity consumption) confirms the model’s high forecasting accuracy and robustness to concept drift. The article details the agent architecture, algorithmic loop, experimental setup, and computational efficiency of the approach. The paper highlights key advantages of the approach, including robustness to concept drift, real-time adaptability, and low computational overhead.

With the rapid advancement of artificial intelligence, and deep learning in particular, models have emerged that are capable of delivering highly accurate predictions. However, the internal logic of such models remains difficult to interpret—an issue of critical importance, especially in domains where the correctness of an algorithm directly affects high-stakes decision-making. One promising avenue for addressing this challenge is Explainable Artificial Intelligence (XAI), which focuses on developing approaches that clarify model behavior and provide transparent reasoning behind the results obtained. This work examines theoretical foundations of XAI, with particular attention to the classification of methods and the challenges posed by the "black box" nature of machine learning models. The review highlights the necessity of advancing new XAI techniques, outlines potential ways to reconcile high predictive accuracy with sufficient interpretability, and lays the groundwork for further research in this field.

The article is devoted to methods and algorithms for constructing a fuzzy model for selecting contour pixels. There are an overview of expert approaches to constructing fuzzy models for selecting contour pixels and noted advantages and disadvantages. To automate the construction of fuzzy models for selecting contour pixels, a method for developing a composition of fuzzy production rules is proposed, based on Tsukamoto’s fuzzy inference and the analysis of brightness gradients of eight neighboring pixels in the direction of the pixel under consideration. To build the model, a single grayscale image with normalized brightness values is used. To decide whether a pixel belongs to a contour, eight linguistic variables of the "brightness gradient" model are introduced for each pixel surrounding the pixel in question, and a linguistic variable "contour belonging" is also introduced. For each linguistic variable introduced fuzzy sets with parametric membership functions. To construct the structure and composition of a fuzzy model, proposed an approach based on the correlation of a halftone and its black-and-white equivalent. To optimize the parameters of the membership functions used a data set which is formed on the basis of the original halftone image and the contours of objects applied to it. The generated data set is used by a genetic optimization algorithm. For each chromosome of the population, the fitness function is calculated using a fuzzy model with the corresponding parameter values. Given examples of application of the developed fuzzy model to other images.

Presented is the development of an operator-oriented user interface (UI) for visualization and control of territory surveys using an unmanned aerial vehicle (UAV) equipped with ground-penetrating radar (GPR) and additional sensor subsystems (magnetometer, metal detector, optical camera, inertial navigation system (INS)). The solution targets humanitarian demining and subsurface mapping tasks. The interface implements the full operator workflow — "route planning → execution → archive/reporting → settings/diagnostics" — supports geo-referenced multichannel data (B-scans — profile radargrams, magnetic and induction profiles, photologging), and stores results in a normalized relational database (DB). The system architecture is oriented to the GPR signal chain ("antennas → receiver → digitization → processing → integration") with time and position synchronization (GPS/GLONASS). Tools are provided for generating unified reporting materials: a track map, a table of detected objects (type, depth, coordinates, confidence level), and attachments with raw radargrams and images. It is shown that an operator-centered interface reduces cognitive load and improves process observability through structured visualization and consistent interaction scenarios. This enables reproducible operational evaluation and preparation for field trials. The approach is consistent with current practices of integrating GPR + UAV and multisensor complexes in applied object detection and classification tasks [1—4, 11—12].