Abstract
We believe that robotics can save a lot of money and time for everything related to urban agriculture. Having already been sensitized on this subject thanks to the vegetable garden on the roof of a building of our school (Eiffel 1), we thought of a project which would help the assistance of "urban vegetable gardens". We want to create a robot that would take care of this type of space in a green, autonomous and safe way. He should be able to:
- Travel independently in a relatively rugged environment
- Map its environment
- Recognize the weeds present in the vegetable patch
On the other hand, we want it to be able to adapt to all vegetable garden sizes; allowing to create a product adapting to the needs of all, whatever their spaces. So, one of the future visions that we have for this PFE would be that it could respond to these same problems on a larger scale - than a private garden. This is why we want to focus
on a fairly open issue which is:
How to make a plant recognition system adaptable to all gardens?
With this project we are therefore targeting an informed audience in electronics (loop systems, mechanics, robotics in general) and who could therefore understand and improve the functionality of our project. There are many communities that may be interested in this kind of project; for example, Fablabs, "makers" or even educational infrastructures.
We want to create a base that everyone can understand and then allow those who have the skills to modify our project.
We managed to make a robot using the ROS architecture in order to ensure that it meets the needs mentioned above. This one:
- Moves and potentially autonomously thanks to mapping (LIDAR) in a relatively rugged environment
- Mapping its environment
- Detects objects (and potentially weeds in the garden)
This project is promoting open innovation because we start from existing technologies as a source (such as the pre-trained neural segmentation network, weed databases or
even robot implementations under the ROS platform ...) and that we would like to make our contribution to the building by grouping these functionalities on a specific theme and adding new ones. theme and adding new ones.
On the other hand, we also have a lot of ideas about the future of this project and we found it relevant to make it (in open innovation) available to students of the school (or any other person curious about robotics and of its evolution). These future perspectives are:
- Ability to remove unwanted plants with a robotic arm
- Identify different plants (not just weeds)
- Program fully embedded AI (without remote PC - because currently our calculations of plant identification by neural network and those concerning the autonomous movement of the robot are carried out on a PC with ROS, Linux & TensorFlow and communicating by Wifi with the Raspberry).
- Improved mapping, ability to set points of interest manually.