Improving Greenhouse Microclimate Control Efficiency Using Neural Network Technologies and Distributed Embedded Systems
Abstract
Improving Greenhouse Microclimate Control Efficiency Using Neural Network Technologies and Distributed Embedded Systems
Incoming article date: 10.05.2025This paper explores the development and application of a neuro-inspired automated system for greenhouse microclimate control. The relevance of the topic stems from the need to improve the resilience of agricultural production in the context of population growth and climate change. The proposed system architecture includes a hierarchical sensor network based on STM32 microcontrollers, an analysis module implemented on a Raspberry Pi 5 with neural network-based data processing, and a distributed actuator layer. The paper describes telemetry processing methods, device addressing and polling algorithms, as well as neural network models (YOLO and EfficientNetB3) used for plant disease diagnostics. An experimental evaluation of the classification model demonstrated high accuracy and confirmed the system’s capability for real-time operation.
Keywords: greenhouse automation, microclimate, neural network analysis, STM32, Raspberry Pi, deep learning, YOLO, EfficientNet, IoT, plant disease classification