Description
Norway has a sparse distribution of fertile soils and high labour costs, which favours sustainable farming with high-value crops. We argue that such farming can be made more efficient, and resilient to climate changes, through the integration of flexible and low-cost robotic technologies. We propose a case study on partially automated sugar pea harvesting in Norway. The pea has a high protein content and, as a leguminous crop, a low demand on nitrogen fertilizer - an environmentaly friendly crop. However, the harvesting is highly labour-intensive. The project goals are: a) to perform a systems analysis of the harvesting, identifying needs and requirements; b) to develop a mathematical moldel of the adaptative robotic system(s) required to replace manual harvesting, and c) to implement and test in-situ a first proof-of-concept robot picker. The multidisciplinary team working with the applicant includes expertise in advanced robotics (Spain), precision agriculture (in particular remote and proxymal crop sensing) and sugar pea cropping (Norway).
Summary of project results
The horticultural sector is characterized by a large share of manual tasks. Flexible and low-cost robotic technologies hold promise for overcoming high labour costs. The market for sugar peas in Norway is growing rapidly. The objective of the work presented was a low-cost robotic system for picking flied-grown sugar peas, which can be operated in a fleet, and where the fleet size can be scaled to the need of the farmer. It is believed that such a system could help producers in industrialized countries with high labour costs reduce the reliance on imported seasonal manual labour. This project brought together robotics expertise from Spain and precision agriculture, and horticulture expertise from Norway. It was hoped that this could be the bases for a longer-term collaboration towards such robotics systems. The Project stay started with a requirements analysis of the problem at hand, investigating the potential for low-cost robotic picking of sugar snap peas. A first field-study on a real end-user farm, Torbjørnrød Gård, was also performed. Here the problem was further discussed with the farmer, and the Researcher was given an insight into the current manual picking process. Then a suitable off-the-shelf robot arm was selected based on the requirements of the tasks and purchased. The robot arm was then integrated with different sensors, such as a camera and a depth sensor. A mobile base was also selected and purchased, a small 4-wheel drive platform. The integration of arm on the base was begun, to get a first physical prototype. During the Project there was significant effort put into developing algorithms for the segmentation of the sugar peas from the leaves and background. This led to a first simple implementation of autonomous visual servoing to sugar peas by the robot. The testing was performed on real sugar-pea plants in greenhouse conditions at Bioforsk (grown specifically for this Project). A second field-study was performed at Torbjørnrød Gård with the pre-prototype robot. The mobility of the robot was tested and found satisfactory. Manual control of the robot in a simulated picking task was also performed, but autonomous operation was difficult due to the highly variable lighting conditions. This is part of the future work of the Project.
Summary of bilateral results
The partnership of robotics expertise in Spain and agricultural expertise in Norway was essential for the success of the project, given the multi-disciplinary topic. The project has included two meetings with a relevant end-user, Torbjornrod Gard, one of the largest sugar pea producers in Norway, A one-day mini-workshop with the personnel involved from Bioforsk and Gartnerhallen was also organized, with a tour of the facilities at Bioforsk, demos of the robotic systems, and meetings, Oral presentations and papers are in preparation at the closure of the stay. The project has created a bond between Bioforsk and Universidad Carlos III de Madrid, and the collaboration is being continued through an extension of the project. A close collaboration with Garnerhallen, a large agricultural collective, has been mantained throughout the project. The extension of the project has also attracted support from Bama (the largest private distributor of fruit and vegetables from Norway). The project has benefited the host institution in Norway, Bioforsk, thorough the collaboration with Universidad Carlos III de Madrid in Spain, through publications and publicity, and through the robot hardware and software being built for the project. It has also benefited Gartnerhallen SA, which would like to use the research to help their partner members that are growing sugar peas. Finally, the project has benefited Universidad Carlos III de Madrid with publications and with the collaboration on the Master thesis of a student. At the closure of the stay, there is an ongoing discussion among partners about applying for future collaborative research projects on agricultural robotics, and picking robots for horticulture in general.