AI Farm
Smart Horticulture Automation Lab IoT AI Innovation Lab
🌿 Project Overview
AI Farm is a small-scale experimental horticulture site in Belgium. The goal is to reactivate and modernise an existing growing automation setup, and evolve it into a practical testbed for AI-driven cultivation, IoT sensor networks, and small-scale robotics — all within the context of a real horticultural land use.
- Location: Belgium
- Focus: Practical AI and automation experiments in horticulture
- Scale: Small-scale, iterative, and fully hands-on
- Approach: Real sensors, real crops, real data — no simulations
- Public presence: Dedicated website communicating activities and land use
🧪 Experiment Areas
Each area is a standalone experiment that can be built and tested independently.
🌡️ Climate Monitoring
Temperature, humidity, CO2, and light sensors distributed across growing zones. Real-time dashboards and historical trend analysis to understand microclimate behaviour.
💧 Automated Irrigation
Soil moisture sensors combined with automated valve control. Rule-based irrigation schedules with AI override based on weather forecasts and growth stage.
📷 Growth Monitoring
Camera-based plant monitoring using computer vision to track growth stages, detect anomalies (pests, disease, nutrient deficiencies), and estimate harvest readiness.
🤖 Robotics & Movement
Small autonomous platforms for inspection and light tasks. Navigating rows, capturing sensor readings, and eventually assisting with seeding or harvesting.
📊 AI Decision Support
Aggregating sensor data, images, and external signals (weather, market) into an AI agent that suggests actions: when to water, when to harvest, when to intervene.
🌐 Public Website
A clean, professional public-facing website for AI Farm that documents activities, communicates the horticultural land use, and positions the site as a credible innovation project.
🏗️ Technology Stack
A pragmatic mix of proven IoT building blocks and modern AI tooling.
Edge & Sensors
Connectivity & Data
AI & Intelligence
Automation & Control
⚙️ How It Connects
All components talk to a central edge hub; the AI layer sits on top as an optional decision layer.
ESP32 nodes measure climate, soil, and light across zones. Data is published over MQTT to the local hub.
Raspberry Pi runs Node-RED for data routing, InfluxDB for time-series storage, and Grafana for visualisation.
Rule-based triggers drive relays and valves. Schedules can be overridden by the AI decision layer in real time.
An LLM agent aggregates sensor history, camera observations, and external data to provide recommendations and autonomous actions.
A dedicated website surfaces live status, experiment logs, and a transparent account of activities on the site.
🗺️ Roadmap
Phase 1 — Foundation
Reactivate the site; install climate and soil sensors; set up MQTT hub, InfluxDB, and Grafana; launch the public website.
Phase 2 — Automation
Add irrigation control, relay boards, and scheduling. First AI experiments: rule-based decisions backed by sensor data.
Phase 3 — Vision
Introduce camera-based growth monitoring and computer vision models for anomaly detection and harvest stage recognition.
Phase 4 — Robotics
Deploy a small autonomous inspection platform. Integrate an LLM agent as the central decision layer across all inputs.
🎓 What AI Farm Demonstrates
AI Farm is where software engineering meets the physical world:
- ✓ IoT in practice — Real sensors, real data, real constraints
- ✓ AI beyond the screen — LLM agents controlling physical actuators
- ✓ Edge computing — Offline-capable systems running on low-power hardware
- ✓ Iterative innovation — Each phase delivers something tangible and testable
- ✓ Responsible land use — Technology in service of legitimate horticultural activity
- ✓ Public transparency — Open communication about what is being built and why