CybeleTech is a French SME, established in 2011, that aims at developing the use of numerical technologies in agriculture. The core products of CybeleTech are based on numerical simulations of plant growth through dedicated biophysical models and machine learning methods extracting knowledge on processes through large databases. These new technologies can bring added values at different stages of the agriculture and food chain: plant breeding, optimization of cultural practices, forecasting yields and production at large scale for insurance, optimization of first transformation processes. These algorithms and the associated data are resource consuming, in terms of computation and storage.
In Physics CybeleTech is leading the smart-agriculture use case dedicated to optimization of deployment of crop monitoring for decision aid solutions on Edge for crops needing precise and quasi real time management like greenhouses. To achieve better use of theses complex cultivation system, growers need to anticipate how the setup of the greenhouse devices, such as heating, CO2 generator or irrigation system, will impact plant development and then the operator’s gross margin.
As depicted in the image below, the decision support system developed by Cybeletech enables to connect to the greenhouse for collecting data measured by various sensors. These data are transferred to Cybeletech servers where there are pre-processed and stored in appropriate databases. Agronomic models are fed with these data, allowing to explore different management scenarios in term of production cost and outcome. Furthermore, to ensure realistic crop growth simulation, agronomic models’ parameters must be adjusted through calibration procedure with high computational cost.
The main challenges faces are: 1) To ensure a continuous flow of data between greenhouse and the databases with high resilience to connection failure; 2) To ease the deployment of the DSS in very different and specific infrastructures; 3) To improve the computational efficiency for optimization and model calibration procedures, while reducing the costs for growers.
Edge computing will allow to store the environmental data collected by the sensors at the most local level, thereby limiting the risk of data loss due to unstable internet connection. Moreover, it will allow to run short-term simulation based on these local data directly in the greenhouse ensuring autonomous, near real-time greenhouse management optimization. On the other hand, cloud computing and FaaS will allow to improve computational efficiency, while reducing the costs for growers.
The PHYSICS project will provide precious tools to facilitate the uptake of those two architectures and ensure the continuity between them.
