Increasing intelligence through machine learning
In light of the large number of different plants, Bonirob’s automatic image recognition plays a key role. Albert describes the challenge: “The leaves of carrots and chamomile, for example, are very similar in their early stages.” As a result, he has to teach Bonirob how to learn and recognize the shapes of leaves. How do you “explain” the shape of a carrot leaf to a robotic system? Albert and his team use what is known as machine learning. This involves a large number of image files in which the Bosch researchers highlight the weeds. “Over time, based on parameters such as leaf color, shape, and size, Bonirob learns how to differentiate more and more accurately between the plants we want and the plants we don’t want,” Albert says.
New business thanks to agile teams
Albert and his team are developing the agricultural robot at Deepfield Robotics, a Bosch-owned start-up company that emerged from the work of a corporate research team in 2014. Bonirob is the product of a public joint project funded by Germany’s Federal Ministry of Food and Agriculture that saw experts from Bosch, Osnabrück University of Applied Sciences, and the agricultural machinery manufacturer Amazone join forces. Under the auspices of Robert Bosch Start-Up GmbH, Bosch has since taken over the task of further developing this high-tech tool. At the 2015 European Robotics Forum in Vienna last spring, Bonirob was singled out for a 2015 euRobotics Technology Transfer Award. In September, the German Federal Minister of Food and Agriculture Christian Schmidt presented the agricultural robot with an award for innovation in horticulture. It won the Deutscher Innovationspreis Gartenbau in the “technology” category.