From sensor data to real-time agricultural decisions (handling unreliable data)

Most AgTech demos look like this:

• clean historical datasets
• controlled environments
• perfect data

Reality is different.

Sensors fail.

Data is noisy.

APIs go down.

And the system still needs to decide.

🌱 What I’m building

AgroSentinel is a real-time agricultural decision system.

It connects IoT sensor data with:

• irrigation recommendations
• plant stress detection (VPD-based)
• phytosanitary risk evaluation

👉 live system:

https://agrosentinel.dev

⚙️ Architecture

ESP32 → ingestion API → data validation → decision engine → Orion Context Broker → dashboard

🔧 Example (NGSI update via Orion)

bash
curl -X POST http://<orion-host>:1026/v2/entities \
  -H "Content-Type: application/json" \
  -d '{
    "id": "WeatherObserved:FieldSensor",
    "type": "WeatherObserved",
    "temperature": { "value": 22.5, "type": "Number" },
    "humidity": { "value": 65 }
  }'