Abstract
Smart agriculture systems increasingly adopt edge–cloud fusion to improve the responsiveness and reduce the data overload. However, many existing models still depend on static or centralized transmission, resulting in high latency and inefficient communication. This paper proposes a Modular Intelligent Resource Architecture (MIRA) that integrates a three-layer fusion model to optimize Quality of Service (QoS) in greenhouse monitoring. The system applies Probabilistic Neural Networks (PNN) at the edge for anomaly filtering and Adaptive Neuro-Fuzzy Inference System (ANFIS) at the cloud for contextual aggregation. Using Network Simulator-3 (NS-3) simulations with 50–300 virtual sensor nodes, three transmission scenarios, namely serial, parallel, and modular, were evaluated. The results show that the modular model significantly reduces the end-to-end delay (35–54 ms), latency (32–50 ms), and improves the response time and Packet Delivery Ratio (PDR) (up to 95%) compared to conventional approaches. The proposed architecture demonstrates superior scalability and reliability under dense network conditions. Its modular design enables efficient, real-time data processing and selective communication, making it well-suited for precision agriculture in resource-constrained environments.

This publication has 23 references indexed in Scilit: