Background. Indexing is a key tool for improving the performance of relational databases, but manual index selection remains complex and subjective. As data volumes grow, the demands on the adaptability of optimization systems increase. Machine and deep learning-based approaches provide high accuracy, but require significant resources and have a “cold start” problem, which necessitates autonomous solutions with balanced accuracy and complexity characteristics. Objective. Development and experimental evaluation of an autonomous index recommendation system that combines a cost-model with a Plan Covering algorithm to improve query performance in relational databases while simultaneously taking into account the cost of index storage. Methods. The system has a pipeline architecture and includes a workload analyzer, a cost model, and a Plan Covering algorithm. The analyzer generates a query profile, the cost model evaluates their performance with hypothetical indexes, and the algorithm performs a greedy index selection based on the benefit-to-cost ratio. Experimental testing was performed on a synthetic e-commerce system dataset in the PostgreSQL environment. Results. The system generated 7 index recommendations, of which 6 were successfully created. The average query execution time improvement was 22.54%, with the largest improvement (26.06%) being obtained for the most frequently executed query with the JOIN operation. The total execution time of all queries decreased from 0.2751 s to 0.2131 s. Conclusions. The proposed combination of Cost-Model and Plan Covering provides effective automation of the indexing process without the use of resource-intensive neural network approaches, demonstrating a significant improvement in performance and rational use of storage resources.
Background. Small and medium-sized businesses in Ukraine operate under two simultaneous pressures: a prolonged martial law with strikes on energy infrastructure, and the need for digital transformation to remain competitive. Existing ERP and CRM systems assume stable infrastructure and large implementation budgets and are poorly suited to partial-availability operation, fast migration between on-premises and cloud environments, and phased deployment within tight budgets. This creates a need for architectural solutions for the class of small multi-channel retail-service businesses. Objective. The paper aims to develop the architecture and algorithms of a modular digital automation system for the multi-channel retail-service small business class under martial law conditions, and to propose a verification and validation methodology for the microservice implementation of its components. Methods. The architecture follows the principle of one microservice per business function and distributes responsibilities among nine services. Service interactions go through standardised APIs (REST, gRPC) and asynchronous messaging (RabbitMQ). The verification and validation methodology covers unit, contract, integration, end-to-end, and chaos testing. Validation was carried out on four enterprise types: retail with service support, e-commerce, food service with catering, and a full production-and-sales model. Results. The architecture supports partial-availability operation when individual services fail, portability between on-premises and cloud infrastructure, phased deployment within tight budgets, and the absence of vendor lock-in due to the open-source stack. The proposed verification and validation methodology covers five testing levels and ties each architectural concern to a specific check type. Conclusions. The modular microservice architecture is suitable for the class of multi-channel retail-service small businesses operating under unstable infrastructure conditions. Further work covers the collection of quantitative operational metrics and the integration of machine-learning-based forecasting modules.
Issues. Mobile wheeled robots are widely used today in internal logistics environments of various industries, in particular in warehouses, production shops and sorting centers. However, in such environments, the robot operates under conditions of high variability, where the configuration of aisles, the placement of loads and the presence of people can change in real time. Therefore, the development of autonomous navigation of the robot in complex and dynamic environments is becoming one of the key tasks of robotics. Research objective. The method of this work increases the efficiency of navigation of a four-wheeled autonomous mobile robot without maps (mapless) in dynamic logistics environments by applying adaptive machine learning with reinforcement (Reinforcement Learning, RL) based on the curriculum. Methods. To achieve the set goal and solve the tasks of autonomous navigation in logistics environments, a comprehensive methodological approach was used, based on the following methods: - simulation method: Using the PyBullet physical engine to create a virtual logistics environment. This allowed for safe and high-repetition training, simulating the real-world kinematics of the Husky four-wheeled robot; - deep Reinforcement Learning Methods: Application and comparison of three state-of-the-art architectures: SAC (Soft Actor-Critic), PPO (Proximal Policy Optimization), TD3 (Twin Delayed DDPG); - adaptive Curriculum Learning (ACL): A step-by-step learning technique that automatically changes the complexity of the environment (number and speed of obstacles) depending on the current success of the agent. This avoids the degradation policy at early stages. Results. The combination of adaptive learning (ACL) of the Husky mobile robot in complex dynamic environments and hybrid reward functions increased the navigation success rate by 35–40% compared to direct learning in complex scenarios. The conducted comparative analysis of machine learning algorithms demonstrated the highest adaptability to non-stationary conditions and the best efficiency of using experience is the Soft Actor-Critic (SAC) algorithm, which achieves success in 89% of cases of application with a minimum number of collisions. Conclusions. A comprehensive approach to map-free navigation of the autonomous mobile robot Husky in complex dynamic environments is proposed, based on a combination of adaptive learning and a physically interpreted hybrid reward function. Experiments conducted in the PyBullet simulator confirmed that a stepwise and adaptive increase in the complexity of the environment stabilizes the learning process of modern deep reinforcement learning algorithms (PPO, SAC, and TD3).
Background. In fusion facilities, the transmission of measurement signals from sensors takes place under conditions of intense ionising radiation, high temperatures and significant electromagnetic interference, leading to the degradation of sensors and transmission lines, increased signal loss and a reduction in the signal-to-noise ratio. The transmission of weak signals over long distances is particularly critical, requiring a well-founded choice of transmission path structure based on a systematic consideration of the characteristics of the sensors and transmission lines, taking into account a range of influencing factors, including radiation exposure, temperature conditions, electromagnetic interference, signal loss and changes in the signal-to-noise ratio. Objective. The paper aims to justify an approach to selecting the transmission path structure for long data transmission lines in thermonuclear facilities based on a systematic and multi-criteria analysis of sensor and transmission channel characteristics. Methods. A systematic analysis of modern sensor solutions and transmission line types was employed, alongside a multi-criteria approach taking into account radiation immunity, temperature stability, noise immunity, signal-to-noise ratio and long-term parameter stability. The transmission path is considered as a set of interrelated elements: the sensor, local converter, transmission line and data acquisition module, which determine the efficiency of signal transmission. To compare alternatives, the characteristics were normalised and visualised using polar diagrams. Results. It has been established that inductive sensors are characterised by high radiation resistance but are limited in measuring static fields, whereas Hall sensors provide direct measurement of static and alternating magnetic fields, yet are subject to radiation-induced drift. It has been shown that bismuth and ceramic Hall sensors exhibit enhanced long-term stability. It has been determined that coaxial cables are susceptible to radiation ageing and signal loss, whereas fibre-optic and mineral-insulated lines provide better noise immunity and parameter stability. The feasibility of local signal digitisation and the use of optical transmission channels is justified. Based on the analysis carried out, the structure of the transmission path has been established, whilst the merits of local signal digitisation and the use of fibre-optic transmission channels to improve reliability and noise immunity have been demonstrated. Conclusions. An approach to selecting the structure of the transmission path is proposed, based on a comprehensive consideration of the characteristics of sensors and transmission lines. It is shown that the most effective solutions for constructing the transmission path are those involving a combination of radiation-resistant sensors, local signal processing and digitisation directly at the sensor, and the use of fibre-optic transmission channels. This approach ensures improved measurement accuracy and data transmission reliability in the long transmission lines of thermonuclear facilities under conditions of ionising radiation, high temperatures and electromagnetic interference.
Background. Sign language recognition using Artificial Intelligence (AI) is a relevant and still unsolved problem. For Ukrainian Sign Language (USL), the task is complicated by a lack of publicly available datasets. Consequently, researchers operating under data-constrained conditions should apply additional features for improving model accuracy. Hand configuration is a fundamental characteristic of lexical units in any sign language. In USL 40 configurations exist. The ability to accurately identify configuration would provide relevant features for AI-based sign language recognition systems. Objective. To develop a machine learning model capable of frame-by-frame recognition of USL hand configurations on a video stream with an accuracy of at least 95% on the test set and an average inference time not exceeding 70 ms. Methods. As input data for the USL hand configurations recognition models, a system of mathematical indicators, calculated from the keypoints coordinates, extracted via MediaPipe pose estimation system, was proposed. The training dataset was created from camera images of USL hand configurations performed by middle-aged subjects. Following keypoints extraction through MediaPipe, the corresponding vectors of mathematical indicators were computed. Various architectures of random forest and perceptron were evaluated. Results. The highest performance was achieved by the perceptron with two hidden layers of 300 neurons each, which takes the vector of proposed mathematical indicators as input. The model reached an accuracy of 98.18% on the test set. During testing on a video stream, it identified without error 32 out of 40 USL configurations, and with only sporadic deviations. The average inference latency was 46 ms. Notably, a perceptron of identical architecture trained on raw MediaPipe keypoint coordinates exhibited lower accuracy on a video stream. Furthermore, the proposed indicators improved random forest model performance, confirming their universality. Conclusion. The optimal model among the considered architectures was a perceptron with two hidden layers of 300 neurons each, which takes the vector of proposed mathematical indicators as input. It identifies in real-time USL hand configurations, achieving an accuracy of 98.18% on the test set. In experiments based on the proposed expert assessment criterion on the video stream, it correctly identified 35% more configurations (80% vs. 45%), compared to the model based on raw keypoint coordinates. Therefore, implementing additional verification based on expert assessment in addition to the standard accuracy metric on the test data is effective. The presented architecture is the first artificial intelligence-based model for the USL hand configuration recognition. Future research will focus on integrating this network into a system for isolated recognition of USL.
