Agriculture represents a consumption of nearly 70% of the world's water consumption. For several years, many intelligent systems have been developed to optimize the irrigation of crops and plants in general.
However, consideration of plant growth is almost never taken into account, and thus smart irrigation strategies are rarely evaluated. Moreover, we noticed a significant lack of databases informing us about the parameters influencing the plant growth parameters.
It is in this context that we wanted to develop an intelligent monitoring system to fill the gaps identified.
The main functionnalities of our system :
- Monitor the growth of the plant, water consumption and and external parameters (luminosity, temperature, soil humidity).
- Control at a distance the watering or define some strategic rules to automate plant watering. This functionality is not necessarily an end in itself, but allows to build a relevant dataset that can be used to train AI algorithms
- Predict the amount of water to deliver to the plant to achieve a certain growth rate over a given period for a given plant. Data used to train the model are data previously collected.
What is our system made of?
Our system is composed of a set of sensors and actuators implemented on an ESP8266, a raspberry that acts as a gateway, a server and an application.
Sensors used :
- Luminosity
- Temperature
- Soil humidity
- Height (ultrasound)
Actuator :
- Solenoid valve
How does it work ?
A platform is placed near to a plant to collect data on its environement with a set of sensors. An ultrasound sensor is placed under a reference point, a leaf for example, to measure the growth evolution. All those data are collected with the ESP and a arduino script.
What's next ?
- Study the relevance of the sensors, and add others if necessary (soil pH etc...)
- Complexify the watering strategy (combine several parameters, not just a threshold value etc...)
- Deploy the system to collect Big Data and doing analysis (improve AI algorithms with features engineering, study the most releavant parameters etc...)
- Complexify AI algorithms. For the moment just water quantity to deliver over 1 hour is predicted. To go further we need to leave the prediction window free, and also predict how to deliver the predicted amount of water. If this amount should be delivered in several times, ? regular frequency ? when in the day ? etc...?-
- Implement a database and not just text files
What are potential applications ?
Our system requires a large amount of data, especially since there is a large variety of plants with different characteristics. This is why a platform, or even a network, can be created to exchange the data collected. This network can concern individuals as well as farmers. For example, if a farmer wants to start growing corn and manage his water consumption intelligently, he can see on the network if data are not already available and exchange them with data that belong to him and can be used by another farmer.
Our system can also be used for the management of green spaces in cities. As the platform is located with its latitude and longitude, we can observe which type of plant is the most adapted to such or such place of the city. That is to say, where to place a particular type of plant in the city so that it does best without requiring excessive water consumption.
Comments