Optimal Pesticide Use
To provide farmers with a personalised AI model for food-safety-optimal pesticide use.
Pesticide poisoning is a frequent cause of sickness both for consumers and farm workers, with up to 44% of farm workers poisoned by pesticides each year.
Although farm pesticides must satisfy specific EU requirements, the allowed produce as well as the usage regulations differ across EU, depending on the climate specifics of each country. Αpart from knowledge of the regulations, farmers need assistance in making sure that their production satisfies, in practice, the limits set by law. This requires a solution that correlates best practice activities, meteorological indicators, and IoT data coming from crop proximal sensing and food processing equipment.
These recommendations can lead to a lower pesticide usage than the maximum residue limits set by the regulatory authorities, which are indicative and cannot in practice take into account the micro-climatological and other business practices of individual farms or regions.
Τhis Use Case will deploy and train AI models, using the EFRA learning approach, for personalised optimal pesticide use recommendations by leveraging both aggregated data from at-the-farm sensors and publicly available data sources for food-safety optimal use of pesticides.
In this use-case we will focus on food-safety-optimal pesticide usage for 3 EU countries. We will deploy and train appropriate AI models using the EFRA privacy-preserving two-level training approach .
At the first level, an AI model is trained on the public and private aggregated data of AGRIVI to provide them with a personalised AI model for food-safety-optimal pesticide use.
At the second level, 3 farmers will be engaged to train a single AI model using a federated learning approach.
To incorporate the at-farm sensor readings, EFRA will develop re-usable cloudlet applications that will run on the local farm premises, pre-process the sensor readings, and deliver higher quality, consolidated results at a slower velocity, such that they can be directly mapped to the AI model parameters.
EFRA will aggregate all the above-mentioned sources of data (both public and private) related to pesticide usage to deploy and train AI models for personalised recommendations on optimal pesticide usage.