Repository containing other repositories related to the greenhouse digital twin project
Digital twins have emerged as a promising technology that enables virtual replicas of physical assets, allowing for real-time monitoring, analysis, and simulation. These virtual replicas can be applied across various fields, including agriculture, manufacturing, healthcare, and more. In this research project, our focus is on building a digital twin for a greenhouse as an example to showcase the capabilities of this technology.
Our approach involves developing a digital twin for a greenhouse using a combination of Python programming, SMOL language, and Raspberry Pi.
The Python program is designed to interact with InfluxDB, a time-series database, to collect and store sensor data, while the SMOL language is used to create a representation of the assets and a knowledge graph that captures the relationships between different components of the greenhouse. The sensors, connected to various Raspberry Pi, are set up to collect data on various environmental parameters such as temperature, humidity, light, and soil moisture, providing a rich dataset for analysis.
The primary objective of this research project is to create a functional example of a digital twin that can showcase its potential in monitoring and managing a greenhouse environment. By creating a virtual replica of the greenhouse, we can simulate and analyze its behavior, and gain insights into how different components interact with each other. This digital twin can serve as a valuable tool for optimizing greenhouse operations, improving resource utilization, and enhancing overall crop yield.
In this report, we will provide a detailed overview of the methodology used to develop the digital twin, present the results of our data analysis, discuss the implications of our findings, and highlight the potential applications of digital twins in various fields beyond agriculture. This research contributes to the growing body of knowledge on digital twins and serves as a practical example of their application in a real-world setting.
InfluxDB is a time-series database that is used to store the data collected by the data collectors.
It is composed of buckets which contain measurements. Each measurement is composed by:
- measurement name
- tags
- Used as keys by us to identify the data
- fields
- Used by us to store the actual data from sensors
- timestamp
and represents a single data point at a specific time.
This particular structure allows getting data for a specific time range and performing aggregations on it (e.g. mean)
The Python influxdb-client is a Python library that is used to interact with the InfluxDB database.
It is used by the data collectors to send data to the influxDB instance running on the host computer.
OWL is a knowledge representation language that is used to represent the asset model of the greenhouse. In other words is used to create a formal representation of the greenhouse physical structure.
The asset model is used to represent the assets described in Assets - Sensors and the relationships between them.
SMOL (Semantic Modeling Object Language) is an object-oriented language that, among others, allows to
- Interact with influxDB to read data from the database
- Read and query a knowledge graph, mapping the data read to objects in the language
- Map the whole program state to a knowledge graph by means of the semantic lifting. The program state can then be queried to extract information
- Represent and run simulations (FMO) and interact with modelica
In our case, it is used to connect the asset model to the data collected by the data collectors, perform the semantic lifting of the program state and interact with simulations to create the digital twin of the greenhouse.
The specific greenhouse we are working with has the following characteristics:
- The greenhouse is divided in two shelves.
- Each shelf is composed of 1 group of plants.
- A single water pump waters each group of plants.
- 2 plants of the same type compose a group of plants.
- Each plant is put inside a pot.
Here is a list of assets we are representing for our architecture, along with the sensors we are using to collect data on them:
- Greenhouse
- Light
- Shelves
- Temperature
- Air Humidity
- Pots
- Soil Moisture
- Plants
- Infrared camera: it calculates the NDVI (Normalized Difference Vegetation Index) of the plant. We will use this data to determine the health of the plant.
- Water pumps
- Water flow
The data collectors are Raspberry Pi devices that collect data from the sensors and send it to the DB. Each data collector is associated to a greenhouse shelf and is responsible for collecting data on the assets that are located on that shelf.
The host computer runs
- An InfluxDB instance that holds data retrieved from the Data Collectors
- A Java program that periodically executes the SMOL Twinning program, which is responsible for creating the digital twin of the greenhouse.
The user can interact with the digital twin though the host computer.
They consist in python scripts that are run on the data collectors and are responsible for collecting data from the sensors and sending them to the influxDB instance on the host computer.
All the scripts are grouped in the greenhouse-data-collector project in the sensors' module. We chose to apply an OO approach to the problem where each measurement (Humidity, Moisture, Temperature...) is represented with a class. When not strictly coupled, physical sensor controllers and measurements are separated from each other, making the code more modular. Each class contains a read() method which returns the value, when needed the class may call an Interpreter to convert the raw value, in that case the mapping can be configured by modifying the config.ini file.
Python program that is run on the data collectors and is responsible for collecting data from the sensors and sending them to the influxDB instance on the host computer.
It achieves this by:
- A virtual representation of the various assets in the greenhouse is created (E.g., Plants in the greenhouse).
- Each virtual asset is connected to a set of physical sensors (E.g., Virtual plants are connected to the appropriate camera that retrieves plant health and growth).
- Virtual assets periodically collect data from the sensors, create a data point containing the asset identifiers and the sensors' detection and send it to the host computer.
The Greenhouse Asset Model is an OWL file representing the physical structure of the greenhouse and the relationships between the various assets in the greenhouse.
The SMOL program uses the asset model individuals as a starting point for twinning the greenhouse.
Here follows a picture of the asset model:
The SMOL program is run by the host computer and is responsible for creating the digital twin of the greenhouse.
It achieves this by:
- Reading the asset model from the OWL file
- Generating SMOL objects from the asset model individuals
- For each asset object, retrieves sensor detections for that specific asset from the influxDB database (E.g., Retrieve moisture data for a specific pot).
- After retrieving the data, the program performs the semantic lifting of the program state, creating a knowledge graph that represents the state of the assets in the greenhouse.
The SMOL program is run periodically by the host computer to retrieve the digital shadow of the greenhouse. A simple Java program is used to schedule the execution of the SMOL program.
Assuming that the host computer and the data collectors are already running as specified in the Project Setup section, the execution flow is the following:
- The data collectors periodically collect data from the sensors and send them to the influxDB database.
- the SMOL Scheduler periodically runs The SMOL program.
- The SMOL program retrieves the asset model from the OWL file and generates SMOL objects from the asset model individuals.
- For each asset object, the SMOL program retrieves sensor detections for that specific asset from the influxDB database.
- After retrieving the data, the SMOL program performs the semantic lifting of the program state, creating a knowledge graph that represents the state of the assets in the greenhouse.
- The SMOL scheduler retrieves the digital shadow of the greenhouse from the SMOL program and sends it to the user.
Check the How to run doc for instructions on how to run the project.