Robotic Crop Phenotyping Testbed for Sustainable Agriculture
To meet the food demands of increasing global population, it is essential to make advancements in the field of genomics and plant phenotyping to explore novel traits in plants. This would require developing automated phenotyping methods which are efficient, and less prone to human error and bias. This robotic platform, equipped with a diverse set of sensors, enables it to gather a rich data set of essential parameters required for digital phenotyping.
Platform
The design of the robotic system is based on 3D printed plates to encompass future expansions and changes to the system as well as modularity for sensor mounts. A cable driven approach is adopted addressing most of the UAV and ground-based vehicles limitations to carry the robotic system.
Sensors
The robotic overhead platform is outfitted with 3 key sensors, as well as a computing module for data pre-processing and temporary storage. These sensors include:
- 3D LiDAR
- 9-DOF Inertial Measurement Unit.
- RTK GPS.
Dataset
The architecture of datasets is significant in assisting researchers and scientists in several fields of research. This data is first captured in the form of .rosbag files comprising sensor data from all of the onboard sensors. These files are later transformed into more diversified and easily accessible forms. The pointcloud map is converted and saved to.pcd,.ply, and.las file formats for use in Python and MATLAB applications. GPS and IMU data are also retrieved from rosbag files and saved in.csv and.txt formats.
Results
The crop's consecutive scans with a robotic platform can disclose a variety of information about cultivar attributes, such as plant growth rate, biomass changes, and diversity within a field. Various phenotypic features of plants in the field may be studied by combining information from multispectral photos taken by UAV.
