A research platform demonstrating how AI systems can shift from reactive responses to anticipatory, autonomous decision-making in agricultural IoT sensor networks.
Traditional AI systems are reactive — they wait for a query, process it, and respond. A proactive AI system anticipates needs, detects emerging conditions, and takes autonomous action before problems occur.
In precision agriculture, this means an AI that doesn't just report "soil moisture is low" when asked, but autonomously monitors sensor networks, predicts drought stress 48 hours ahead, and initiates irrigation — all without human prompting.
Predicts soil conditions before they reach critical thresholds using temporal patterns and weather forecasts. Acts before crop stress occurs.
Makes irrigation, fertilization, and harvesting decisions without human intervention based on multi-sensor fusion and crop growth models.
24/7 IoT sensor network streams data to edge AI processors. No polling, no manual checks — the system is always aware of field state.
Autonomous tractor navigation with pure pursuit control, coverage planning, and obstacle avoidance. The AI's decisions translate to physical action.
Models improve over growing seasons. Soil-specific calibration, variety-specific growth curves, and microclimate adaptation through continuous feedback.
Processing happens at the field level — low latency, works offline, reduces cloud dependency. Critical for rural NC farms with limited connectivity.
Our platform integrates IoT sensors, autonomous navigation, and proactive AI into a unified system for NC sweet potato production.
This research platform demonstrates proactive AI concepts through a functional simulation environment:
Real ESRI satellite imagery with pan/zoom. Geofence drawing, sensor placement, and tractor positioning on actual NC farmland.
Soil moisture, temperature, pH, EC, NPK, leaf wetness, NDVI, weather, water level, O₂, tensiometer, rain gauge, pyranometer, flow meter, CO₂.
Ackermann bicycle-model kinematics, pure pursuit controller, waypoint following, obstacle avoidance, and boustrophedon coverage planning.
Parallel swath guidance with cross-track error display. Headland auto-turns with implement lift simulation.
Rain events, drought conditions, diurnal temperature cycles, GPS noise models (RTK/DGPS/Standalone), and fuel consumption.
OpenAI GPT-4o integration for natural language farm commands, sensor interpretation, and agricultural recommendations.
7-stage growth tracking from transplanting to harvest. Stage-specific soil requirements and AI recommendations.
Real-time tractor telemetry publishing to IoT dashboards. Compatible with Node-RED, Grafana, and cloud platforms.
This work builds on recent advances in proactive AI, agricultural IoT, and autonomous systems:
North Carolina is the #1 sweet potato producing state in the US, with over 95,000 acres harvested annually, primarily in Sampson, Johnston, and Nash counties. The sandy loam soils of the Coastal Plain provide ideal growing conditions, but variable rainfall and soil heterogeneity create challenges that precision agriculture can address.
This project demonstrates how proactive AI can help NC farmers optimize irrigation timing, detect nutrient deficiencies before visual symptoms appear, and coordinate autonomous equipment across large-scale operations — reducing water usage by up to 30% and improving yields through data-driven decision-making.