5G connected spot sprayer enables autonomous control of volunteer potato

Dutch arable farmers face big challenges with unwanted volunteer potatoes. Hot and dry summers result in lower potato yields and more small tubers to be left in the field during harvest. Surviving mild winters, these tubers grow into unwanted potato plants in the next crop in the rotation, like sugar beet or onion, are like a weed and moreover increase soil-borne diseases as potato cyst nematodes or can introduce pests like potato blight. Most effective control method for volunteer potato is the manual application of glyphosate, which requires about 15-20hours/ha of manual labor resulting in costs of €300-€400 per hectare. Decreasing labor availability and increasing labor prices force farmers to look for alternatives. A proof of concept was demonstrated with an autonomous mobile field platform equipped with a spot sprayer and artificial intelligence to recognize potato plants. The spot sprayer consists of four RGB cameras supplemented with artificial light, capturing images of the sugar beet and volunteer potato plants. The images are sent by a 5G connection to a cloud based server about 200 km off site. The server hosts a deep learning algorithm trained to recognize the sugar beet and potato plants in the images. In real-time, the locations of the potato plants are sent back to the spot sprayer in the field. One of the nozzles at a spraying boom with 29 nozzles, spaced 10 cm apart, applies glyphosate on the identified spot of the potato plant. The cycle of capturing images, sending it to an edge server, processing of the images, sending instructions back to the robot in the field and spraying, took 250ms. About 95% of the volunteer potato plants were sprayed and only 4% of the sugar beet were hit. In the presentation we elaborate on the demonstrated proof of concept, the results achieved in the field, advantages of using cloud based edge solutions and discuss future work and perspectives.