Search academic texts

Information × Registration Number 0225U003232, (0122U200452) , R & D reports Title Hardware and software of smart house intelligent systems popup.stage_title Апаратно-програмне забезпечення інтелектуальних систем розумного будинку Head Olenych Ihor B., Доктор фізико-математичних наук Registration Date 03-06-2025 Organization Ivan Franko National University of Lviv popup.description1 The aim of the work is to develop a prototype of a hardware and software system for managing the energy supply of smart objects based on the intelligent analysis of sensor data using machine learning and Fog/Edge computing. popup.description2 A B S T R A C T Report on the scientific research work: 56 pages, 23 figures, 8 tables, 46 sources. ABC-ALGORITHM, FASTAPI, KNN, MLOPS, NVIDIA JETSON NANO, SVM, TF-IDF, EDGE COMPUTING, FEATURE ENGINEERING, TEXT CLASSIFICATION, LOGISTIC REGRESSION, ROUTING, MACHINE LEARNING, STM32 MICROCONTROLLER, NEURAL NETWORKS, NATURAL LANGUAGE PROCESSING, SMART HOME, ARTIFICIAL BEE COLONY. The object of research is a hardware and software complex for intelligent smart home systems, which combines edge computing, machine learning models, and MLOps tools for local data processing and autonomous device management. The aim of the work is to develop a prototype of a hardware and software system for managing the energy supply of smart objects based on the intelligent analysis of sensor data using machine learning and Fog/Edge computing. Research methods – supervised deep learning with feature engineering and quantization of models for prediction and classification; metaheuristic algorithms; experimental and laboratory tests on the created Edge platform. Obtained results. A full-featured MLOps pipeline has been built that automatically collects data, trains, quantizes and deploys models on an STM32 client. A local ABC planner is proposed that successfully finds suboptimal trajectories in a partially unknown environment; flexible weight settings and an inverse problem for their selection (target up to 100% on a simple map) are demonstrated. It has been proven that feature engineering increases the accuracy of detecting fake messages up to 92 %. Combining modern edge approaches and MLOps practices into a single prototype allows you to minimize latency, increase reliability, and ensure independent model updates on local devices. This creates the prerequisites for the widespread implementation of intelligent systems in everyday life and industry.   Product Description popup.authors Boiko Volodymyr Ya. Boiko Yaroslav V. Hura Volodymyr T. Korostenskyi Roman O. Pavlyk Mykhailo R. Prytula Marianna M. Sinkevych Oleh O. popup.nrat_date 2025-06-03 Close