Roots in water, forecasts in data
SmartGreen Terra gathers all of our soilless work, hydroponics and aeroponics. We grow strawberries, basil and vegetables in trays and channels that we measure every day, and use those measurements to train an algorithm that learns to predict how a plant grows.
Why we measure everything
A plant in hydroponics lives inside numbers we can read: conductivity, pH, water temperature, hours of light. If we record those numbers alongside growth, plant by plant, we have the material to teach a model to read the future of a crop.
Two ways to grow without soil
Hydroponics
The roots sit in a solution of water and nutrients that circulates through trays and NFT channels. It is the system we have worked on longest, starting with strawberries.
Aeroponics
The roots hang in the air and receive a mist of nutrient solution. The cabinet prototype is there to find out how much water a plant really needs.
Inside the trays
Roots in view
In the cabinet the roots grow suspended, and you can watch them while they grow.
By hand, flower by flower
There are no insects inside the grow tent: a brush moves the pollen.
The control boxes
Every tray has its own box of probes, and every reading ends up in a log.
The first red
The proof that the loop works is something you eat.
Lit up, behind glass
The case set up in the campus atrium: wood, an LED lamp and the sixteen hole tank waiting for its seedlings.
Wood and chains
A wooden frame, the LED lamp hung from chains and the tank on the table. The plants do the rest.
Work in progress
We wrote the sign by hand: the case is ready, and the tank is waiting for its first seedlings.
The algorithm
The goal is a machine learning model that, from the conditions in the tray, estimates how a plant will grow over the following days. Today we are in the phase that matters most and shows least: collecting clean data, with the same method, for long enough.