AROMA

"Controlling grass cover is an essential factor in agriculture. Weeds compete with the crop and induce yield losses, especially if their development is not managed at early stage. Faced with the gradual disappearance of plant protection products and the appearance of resistant weeds, the use of ""all-mechanical"" technics is the main solution considered to control this factor, in addition to adapted cultivation practices (crop rotation, false sowing, tillage, etc.). For field crops, the current solutions consist of hoeing in the interrow. However, weeds on the row are not destroyed and lead to competition with the existing crop, but also contribute to the multiplication of the seed stock. This is why, as part of the ANR ROSE, the AROMA project (Acquisitions and RObotics for weeding around Corn) aims to develop a mechanical precision weeding machine that works on the row. This complete system consists of a vision sensor, a crop and weed detection algorithm and a servo weeding tool. AROMA combines the expertise of Carré (manufacturer of agricultural machines, the leading French mechanical weeding machine manufacturer), the CEA (organization of basic research in detection and mechanical action and an agricultural technical institute) and ARVALIS (technical agricultural institute of reference for cereals in particular). This partnership, which is deliberately narrowed, ensures expertise on every aspect of this ambitious research and development project. To achieve it, three actions will have to be coordinated: detection, interpretation of data and real-time action, taking into account the operational constraints of the vehicle's forward speed and the final cost of the proposed solution. The AROMA system will be implemented in two cases. Corn, a wide interrow crop, for which no mechanical weed control solution on the row exists on the market today. A lot of development work will be needed not only on algorithms but also on the weeding tool itself. The salad, for which Carré already has a mechanical weeding robot (Anatis). In this case, only the vision and detection part will be improved and evaluated. At the end, the expected results of this project are an inventory of available technologies and equipment, for detecting weeds in a crop and for weeding on the row. Secondly, a complete system from vision to implementation will be developed for use on salad (from the existing robot Anatis) and maize (original development). Finally, an in-field evaluation of the solutions developed will be carried out in situations close to agricultural conditions. This evaluation will focus on weed detection performance, on the row weed control effectiveness and technical-economic acceptability of the developed system."