Behavioral cloning allows creating models capable of making decisions in the same way humans do. Although various models successfully predict human actions in games, their ability to imitate individual players decreases when game variants change. In this work, we explore domain adaptation and personalization strategies using Fischer Random Chess with random piece placement. Using the Maia chess model presented by Tang et al., we tested several fine-tuning strategies. The results revealed three key patterns: old player records have no impact; each new record is highly effective; domain adaptation and personalization complement each other - their combination yields the best result.

A reproducible hybrid approach to learning autonomous control of a robotic device has been proposed and experimentally verified, implemented using publicly available tools without specialized computing hardware. The approach combines learning from operator demonstrations, reinforcement learning, curriculum learning, and domain randomization into a single sequential framework. Based on a comparative analysis of five prototypes representing key classes of architectural solutions, applicable approaches for resource-constrained environments are identified. Three controlled experiments in Unity ML-Agents demonstrated the superiority of the hybrid combination over isolated approaches in terms of both learning speed and robustness to environmental changes. Limitations associated with catastrophic forgetting are outlined, and prospects for integrating large language models to generate synthetic training data are identified.

In open-world games, classical tools of dramatic analysis – the three-act structure, Freytag's pyramid, the hero's journey – fail to describe the actual gameplay experience: players gain freedom to choose the traversal order of modular content, making authorial control over pacing impossible. This paper introduces the concept of storytelling literacy – a player competency encompassing recognition of the emotional register of content, conscious pacing construction through activity sequencing, and narrative framing of one's own gameplay trajectory. Based on a formal model of player narrative state – an adaptation of the story-object/phase/facet/threshold-state formalism – we propose an architecture for an IT system supporting player-driven narrative. The system comprises a Narrative State Engine (feedback loop between action telemetry and the content subsystem), trajectory visualization tools for players, narrative coverage analysis tools for designers, and an adaptive content generation module based on large language models.

The paper addresses the problem of visual representation of temporal and probabilistic elements in formal modeling tools for game mechanics. An analysis of existing visual modeling systems – Machinations, Petri nets, UML state diagrams, and UPPAAL – reveals their key limitations in describing complex game scenarios. An approach to visualizing temporal operators and probabilistic constructs in the Tula tool developed by the authors is proposed. A set of visual primitives is described that ensures bidirectional correspondence between visual representation and formal specification in the gameplay specification language. Examples of visual modeling of typical game scenarios with time delays, conditional probabilities, and multiplayer interactions are given. The possibility of automatic transformation of visual models into formal specifications suitable for verification with PRISM and UPPAAL is demonstrated. A comparative analysis of the expressive capabilities of Tula and predecessor systems is conducted, showing that Tula is the only system among those considered that simultaneously provides formally verifiable semantics, built-in support for temporal operators (X, G, F, U), and arbitrary probability distributions, as well as a visual notation organized around resource flows and game events. The two-level specification principle with bidirectional transformation proposed in Tula closes the design–formalization–verification–correction loop, lowering the entry barrier to formal methods for practitioners without loss of verification completeness. This makes it possible to consider Tula as an intermediate layer between iterative game design and industrial verifiers.

The article addresses the issue of designing a non-violent combat system for a children's RPG game. It presents the results of an empirical research. Its purpose is to obtain data to determine the balance between tactical control and direct damage mechanics.

The scientific novelty of the study lies in the empirical identification of the structure of perception by children aged 6–12 of various types of game mechanics: tactical (indirect impact on the enemy), tactical-force (neutralization without direct damage), and force (direct damage). Based on a combination of card sorting and associative analysis methods, a matrix of similarity of mechanics reflecting their categorization in children's minds was constructed, and stable emotional reactions and associations of children regarding key elements of the combat system (ice – control, wind – humor, electricity – ingenuity, fire – danger, etc.) were identified.

As a result of the study, three clusters of mechanics were identified (“Turn Skip,” “Weakening and Restriction,” “Direct Damage”) and a preference rating was compiled, in which tactical mechanics took the leading positions, while force (direct damage) mechanics were not a priority and were described by children as boring. Thus, the hypothesis regarding the preferences of the children's audience was fully confirmed: the core of a safe combat system should consist of non-violent tactical mechanics (70-80%) that ensure the intellectual engagement of players, combined with tactical-force mechanics aimed at their emotional release.

The practical significance of the study is determined by the possibility of applying the developed recommendations in the design of children's video games, as well as the adaptation of the proposed methodology for similar studies.

A practical approach is proposed for building a specialized corpus of game scripts based on a combination of manual annotation and semi-automatic parsing of the source texts. The method integrates a graphical editor for expert annotation, a modular parser that performs initial structuring of the source texts using formal rules and heuristics, and a visualization of links between scenes. This combination provides an iterative cycle of expert validation necessary to obtain a correct and reproducible corpus.

This article presents a newly developed visual perception algorithm for game agents, implemented in the Unity game engine. The proposed method is based on comparing images from two cameras, taking into account complex visual effects (lighting, shadows, occlusion), and is supplemented by line-of-sight checks, consideration of object velocity, and gradual detection mechanics. The algorithm was optimized using asynchronous computations, dynamic camera activation, and accelerated algorithms. The system was tested under various load levels, and conclusions were drawn regarding the optimal conditions for the algorithm’s operation. The paper also analyzes the scientific literature on similar solutions, identifying their strengths and weaknesses. The results can be applied in video game development to create realistic behavior for non-player characters, especially in games with stealth elements.

This article presents a comprehensive analysis of the application of virtual reality (VR) technologies in medical education, with a particular focus on obstetrics and gynecology. It examines the pedagogical rationale for introducing VR (constructivism, Kolb’s experiential learning theory, cognitive load theory), modern technological platforms with varying levels of immersion, as well as the geography and practice of introducing VR technologies in leading medical universities in Russia and around the world. Particular attention is paid to specialized VR simulators for practicing obstetric care, emergency obstetric conditions (shoulder dystocia, postpartum hemorrhage), and minimally invasive gynecological interventions. The results of randomized controlled studies confirming the effectiveness of VR training in the formation of psychomotor skills, clinical thinking, and team interaction are presented. The main barriers to the implementation of VR are analyzed: financial, technical, personnel, as well as specific problems of cyber sickness and ethical challenges. Organizational approaches to expanding the use of VR are reviewed: models for the collective use of equipment, the development of verified educational content repositories, and integration with learning management systems. The potential for convergence of VR with artificial intelligence (AI) to create adaptive, personalized educational environments is substantiated.

The paper is devoted to the development and experimental verification of a modular user interface architecture for game projects in Unreal Engine. The study identifies typical UI design antipatterns (monolithic widget structures, hard references, lack of abstraction, platform dependency) and provides an analysis of existing architectural approaches (Data-Driven UI, MVC, MVVM), demonstrating their limitations. The authors propose an architecture that integrates the Common UI framework, an asynchronous loading mechanism with soft references, the Gameplay Tags system, and a data registry for dynamic settings menu generation. The architecture is implemented as the AdvancedGameUserInterface plugin. Its experimental evaluation involving 12 developers and 5 UI designers confirmed the effectiveness of the proposed solution across three key metrics: maintainability, performance, and scalability. Based on the research results, practical recommendations are formulated for applying the developed architecture depending on the scale of the game project. The research findings can be used by developers to improve quality and reduce maintenance costs for user interfaces in game projects on Unreal Engine.

This paper presents the development and evaluation of a toolkit designed to optimize the workflow of game designers working with narrative content and game design for indie RPGs in the Godot engine. The relevance of this study stems from the growing number of independent studios utilizing Godot and similar game engines, which necessitates the creation of specialized tools to compensate for the lack of built-in solutions for RPG development, such as dialogue systems, turn-based combat design, cutscene management, and complex user interface implementation.

The aim of the work is to design a toolkit based on an architectural analysis of the reference game Persona 5 Royal, ensuring an optimized workflow for the game designer and reducing their dependence on the programmer. Methodologically, the research relied on identifying player states, independent modules, and areas of professional responsibility, which led to the development of five tools. Their effectiveness was evaluated with the participation of 12 novice game designers, including task completion time measurements, interviews, and expert ranking.

The results of the study showed that tools without analogues in Godot ("Combat Segments," "Complex User Interface," "Dialogue System") received high expert evaluations. The "Interactive Location Objects" tool, which has an analogue in Godot, reduced task completion time by an average of 65.5%. The "Commands to the Game World" tool did not demonstrate effectiveness and requires further improvement.

The practical significance of the research results lies in the creation of a ready-to-implement toolkit that enables indie teams to reduce RPG development time and increase the autonomy of their game designers.

The article considers an approach to the development of an intelligent control system for the illumination of a workspace that corrects the intensity of light based on the analysis of the user's cognitive state. Illumination adaptation is performed in real time using data from the BrainBit EEG Waves neurointerface for electroencephalography (EEG) registration. Based on the spectral analysis of the EEG, indicators of concentration and emotional relaxation are calculated, which are used as control parameters for a smart lamp. The Yeelight LED Bulb W3 has been selected as a smart lamp. Lighting is controlled locally via LAN Control protocol without data transfer to cloud services. Experimental studies have confirmed the operability of the system, low reaction delay and increased subjective comfort of the user. The developed system can be used to increase the effectiveness of the educational process of students.

This article is devoted to the study of the possibilities of the SimInTech simulation environment for creating a software package suitable for identifying dependencies between the parameters of the automatic control system (ACS) and the quality indicators of its operation. The article discusses the tools of the selected environment, based on which a set of educational models has been developed. The article shows the results of the above-mentioned educational complex in the form of solving the problem of correcting a linear continuous automatic control system with negative feedback using a PID controller that is part of the automatic control system, the problem of adaptive control of a linear continuous system with a clearly defined reference model, and the problem of multidimensional scanning to identify the patterns of the selected system

The results of research on parallel applications for heterogeneous architectures using the SYCL standards and the DPC++ language, as well as OpenMP, developed by students of the Department of Applied Mathematics and Artificial Intelligence (AM&AI) at MPEI during research work are presented.

As an applied problem, the search for shortest paths in graphs of various dimensions using the Bellman–Ford and delta‑step algorithms is considered. The impact of algorithm choice on the solution efficiency, as well as the heterogeneous multi-core architecture, is examined.

The Bellman-Ford algorithm is easily parallelized and demonstrated excellent scalability. In contrast, using a GPU for the delta-step algorithm did not provide significant speedup due to the complexity of memory coordination and the need to repeatedly copy data between devices. Testing was conducted using the following hardware: an Intel Core i5-12400F CPU (6 cores, 12 threads) and an NVIDIA GeForce GTX1660 GPU (1408 CUDA cores). The test results show that even within a single computational task, different implementations can exhibit different behavior when the task parameters are changed. This confirms the importance of preliminary analysis and testing when choosing a hardware&software architecture for tasks requiring high performance.

The developed software suite visualizes the dependence of parallel algorithm implementation speedup on the number of vertices in the graph for various method parameters and is successfully used in the educational and research processes of the AM&AI Department at MPEI.

paper considers a domestic web platform by Laboratorium LLC intended for conducting laboratory works on real equipment in a remote format. Access to the laboratory stand is organized through a browser within managed sessions: the student observes the installation via a video stream, performs control actions, and generates an electronic protocol. The instructor is provided with tools for assigning works, scheduling the queue, and verifying protocols against execution logs. Usage scenarios of the platform and options for networked interaction between educational institutions are described. The platform is currently undergoing pilot testing; preliminary results are assessed as positive.

The article presents an ontological model for operator hand gesture recognition based on hand skeletonization using the MediaPipe Hands library and an original method for cognitive transformation of finger states from skeleton keypoints into binary code with its subsequent conversion into a digital control signal. Improvement in the technical and methodological characteristics of the system is achieved through explicit definition of cause-and-effect relationships between objects and processes within the ontology, as well as the application of clear binary encoding rules. The mathematical model of the system includes the following stages: video frame preprocessing, hand skeletonization, coordinate analysis of keypoints, binary encoding of finger states, and generation of the final decimal control signal. The developed gesture recognition interface ensures an accuracy of 98% under normal conditions and no less than 94% under noisy and low-light conditions (50 lux or less), with an average response time of 0.5 seconds, outperforming baseline methods based on widely used libraries such as OpenCV and TensorFlow.

In autoregressive text generation using language models (LLM), the next token is selected using a softmax function with a fixed decoding coefficient equal to one, which is the same for all generation steps and all text types. This approach does not take into account the dynamically changing linguistic characteristics of the generated context. For example, a text with high lexical diversity requires high selectivity, meaning the decoding coefficient should be small. A monotone, repetitive context requires a smoother distribution, meaning the decoding coefficient should be large. This article proposes a fuzzy controller for calculating the decoding selectivity coefficient using two linguistic features: the lexical diversity of the generated context and the Shannon entropy of the probability distribution at the previous step. The fuzzy controller is implemented using the Mamdani algorithm with a base of nine fuzzy logic rules and a defuzzifier based on the center-of-gravity method. The effectiveness of the proposed approach lies in the fact that for terminologically rich text, the fuzzy controller reduces the Shannon entropy of the probability distribution by more than 20 times compared to the standard approach.

This paper proposes a method for adaptive control of sliding window parameters in large language models (LLMs) based on Mamdani fuzzy inference with defuzzification using the high-speed area ratio method (MAR-II). The input variables are the type-token ratio (TTR) and the average BPE token length (ATL), interpreted as measurable correlates of the cognitive load of a text fragment. A nine-rule fuzzy controller generates the sliding window stride and context length (max_length). MAR-II is compared with the center of gravity (CoG) method across a 25×25 grid of input space points (RMSE: 99.9 tokens for stride, 147.4 tokens for max_length). It is demonstrated that MAR-II ensures the additivity of the fuzzy model and eliminates the systematic errors of CoG, achieving a single-inference computation time of approximately 35 ns on a Xilinx Spartan 3E FPGA.

This paper proposes a new approach to determining the level of internet censorship. Our definition of censorship is based on the principle of accessibility of alternative viewpoints in search engine results and large language models. The calculation method involves compiling a list of 11 questions, the answers to which suggest opposite opinions depending on the country. These answers are evaluated in comparison with the government's point of view. For search engine results, calculate the number of answers to these questions from a pro-government point of view and the proportion of answers that detail an alternative point of view. The difference between these indicators is proposed to be considered an index of internet censorship. Since Internet searches are increasingly performed through chatbots of the most well-known large language models (LLMs), the level of censorship is also assessed for these models. The comparison showed that the responses of the national search engine Baidu in China are closest to the government's point of view, while the level of censorship in the United States is higher than in Russia. The effectiveness of censorship in the United States is explained by a different mechanism, whereby materials expressing alternative opinions restrict access to popular information platforms. Censorship in chatbots is much stronger than in internet search engine results are calculated.

A method for integrating the cognitive performance of civil aviation flight and air traffic control personnel has been developed. It is based on the quantitative formalization of eight key cognitive functions: attention, working memory, mental processing speed, spatial abilities, logical thinking, numeracy, control accuracy, and emotional stability. Saaty's analytic hierarchy process was used to determine weighting factors reflecting the relative contribution of each function to the integrated indicator. An expert group of seven specialists from three professional strata was formed (active pilots with over 1000 flight hours, researchers specializing in aviation psychophysiology, and psychologists from the medical flight expert commission). The consistency of individual expert assessments was verified by calculating the consistency ratio (CR), all values of which did not exceed the threshold of 0.10. The consistency of the opinions of the entire expert group was confirmed by the Kendall concordance coefficient (W=0.682) with a high level of statistical significance. Based on the aggregation of expert judgments using the geometric mean averaging method, final weighting coefficients were calculated, which made it possible to form a functional dependence of the integrated indicator of cognitive performance. The obtained results demonstrate the dominant role of attention (p₁=0.246) and emotional stability (p₈=0.248) in the structure of professional reliability of aviation personnel. The developed method creates a methodological basis for the objectification of professional selection procedures, ongoing monitoring of the state of cognitive performance of crews, and decision-making on clearance to perform flight tasks.

The design of curricula in the context of the dynamics of educational standards is shifting from a static set of disciplines to system modeling. This approach should take into account multilevel relationships and the logic of forming cross–cutting competencies. The purpose of the research is to create and test methodological tools for the analysis of educational programs of universities based on network analysis methods. The curriculum is formalized in the form of an undirected graph, where the vertices are disciplines, and the edges are connections through common competencies. The work uses a multidimensional network analysis, including the calculation of centrality measures (degree, betweenness, closeness, eigenvector) and key topological metrics: density, modularity, diameter, and clustering coefficient. The empirical base was made up of data from the curricula of three universities in Omsk. The analysis revealed significant structural and topological differences in the architecture of educational programs due to their departmental affiliation and target training models. Network metrics make it possible to quantify the interdisciplinary connectivity and modularity of programs, which is not available for traditional approaches. The results obtained confirm the high diagnostic potential of network analysis for sound optimization of curricula and verification of compliance with professional competencies.

This article studies the problem of extracting formulas of scientific articles presented in pdf format, as well as converting the selected formulas from pdf to LaTeX format. The extraction of the main formulas of a mathematical article and their subsequent presentation in LaTeX format using the large language model DeepSeek is considered. Extraction of the main results of the papers bases on the paper annotations. Then, using a large language model we search fragments of papers describing the main results and highlighting the main formulas within the found fragments. The solution was evaluated using the metrics of precision, recall, and F-measure; the average values of the F-measure for the test collection was 0.93459 and 0.9294. The MySQL database management system is given as a method for storing the obtained data and the description of the database structure is given in text form: database tables, their attributes, as well as types of stored data

The process of applying Generative Adversarial Networks (GANs) for the synthesis of ionospheric sounding ionograms is considered. This task is relevant in the context of constructing a statistical model of the ionosphere through empirical search. This paper describes the specifics of oblique ionospheric sounding as a subject area, particularly outlining the origin of the data and its physical meaning. The dataset, which includes radio sounding session configurations, environmental parameters, and ionograms, is presented and characterized. A method for converting and preprocessing raw data to form a training corpus is described.

The paper presents the general organization of conditional GANs used in image generation tasks, discusses key modifications to their structure specifically for ionogram synthesis, and provides a schematic demonstration of the architecture. Furthermore, the specifics of loss function selection, based on knowledge of the physical meaning of ionograms, are identified. Based on these considerations, a composite loss function is formulated, combining the discriminator's response with heuristics. Finally, the work provides examples of the results obtained, formulates key existing implementation challenges, and proposes potential solutions.