AgriGraph AI:An Innovative GNN and LLM-Powered IoT System for Intelligent Soil and Crop Health Prediction and Optimizati

Overview

The convergence of Smart Internet of Things (IoT) sensors and Large Language Models (LLMs) presents a novel framework for intelligent soil monitoring and precision agriculture. This study proposes an integrated system in which drones equipped with multi-gas IoT sensors collect daily atmospheric and soil-proximal data, including concentrations of ammonia (NH3), methane (CH4), nitrogen dioxide (NO2), and carbon monoxide (CO). The collected dataset is transmitted to a cloud-based LLM pipeline capable of multimodal reasoning over temporal and spatial data streams to evaluate soil quality, detect potential contamination sources, and forecast degradation patterns. LLMs serve as the interpretive and predictive intelligence of this system, synthesizing raw sensor data with historical and contextual information to generate comprehensive soil health assessments. Beyond conventional analytics, the model provides adaptive recommendations—supported by probabilistic modeling and causal inference—to optimize nutrient balance, irrigation timing, and pollutant mitigation. By systematically refining its predictions through continual learning, the LLM develops a dynamic understanding of each monitored environment, enabling early detection of anomalies and facilitating evidence-based decision-making for sustainable agricultural management. The proposed architecture establishes a closed-loop feedback mechanism through which farmers and agricultural stakeholders receive real-time notifications and prescriptive insights via a mobile application interface. This integration enhances proactive soil stewardship, mitigates environmental risk, and promotes long-term productivity. Ultimately, this research underscores the potential of LLMs as cognitive enablers in precision agriculture, demonstrating how intelligent language models can augment environmental resilience and transform soil health monitoring into a predictive and autonomous process.

Competition history

• CSEF 2026 Environmental Engineering (Track 2) (Senior Division) · Entry S-12-20

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• All Environmental Engineering projects
• CSEF Environmental Engineering projects
• Every CSEF 2026 project