Problem Description
According to the researchers at the University of Bonn, global food production is on track to fall by over 20% by 2050, all while the world population rises to 9.8 billion. Many of the hardships that farmers face in meeting this global demand are due to the absence of a reliable labor force, leading to an increase in the awareness of smart agriculture and the need for harvesting robots.
Unlike other staple crops, such as corn and soybeans, green bell peppers have not successfully been harvested en masse. This issue is due to the highly unstructured way green peppers grow and the dense canopy of leaves surrounding the fruit. Small to mid-sized farmers typically hire temporary laborers to hand-pick green peppers with shears at low wages, often relying on high-school students during summer. Unfortunately, during the harvest season from mid-august to the end of September, there’s a severe labor shortage as the students return to school. Time is of the essence, as there is a two to four-week window of time the pepper can be picked, after which it will get soft and sun scarred.
GetAGrip.AI aims to build a world where robots enable specialty crop farmers with a reliable labor force during harvest seasons. This begins with Peter, the autonomous pepper harvesting robot that provides comfort during the incredibly high turnover labor market. By navigating through rows of pepper crops, identifying peppers, and extracting them before they rot, Peter saves farmers time, money, and stress during their busiest season.
The Use Case
Bob owns a mid-sized farm in Iowa, where he grows various crops, including corn, beans, soybeans, and green peppers. During the harvest season, Bob hires local workers to pick the crops. Unfortunately, labor shortages make it difficult for Bob to find sufficient workers. Bob only could find enough workers to harvest his primary crops, including corn and beans. As a result, many green peppers were not harvested in time and rotted, costing Bob significant amounts of money. Therefore, Bob decided to invest in Peter, the autonomous pepper harvesting robot.
Peter navigates around the farm to identify and extract green peppers from their crops during harvest seasons. To identify peppers, Peter scans the plant to detect the peppers, peduncles, stems, and leaves. The peduncle refers to the stem section attached to the bell pepper. Subsequent uses of the word “stem” will refer to the stems of the crop rather than the peduncle. Peter then ranks the visible peppers in a region based on occlusion and depth to determine the order to pick the peppers. After deciding on the pepper picking order, Peter plans a manipulator trajectory that safely moves his robotic arm to the proper location of the pepper to pick with minimal plant damage. Once the manipulator’s arm reaches the pepper, the end-effector grips the peduncle of the pepper. The cutting mechanism of the end effector is then separately actuated, and the pepper is extracted from the crop. Peter then moves his manipulator arm again to place the pepper in a basket and harvest the next one in sight.
When all the visible peppers in a region are picked, Peter navigates to the next plant and continues to harvest peppers. After a few hours of running the robot, Bill logs into the user interface to check how Peter is doing. He sees some peppers were not picked due to occlusion and other complications. Since he is provided with the pictures and pepper locations, he goes to the farm to pick the remaining peppers. Bill collects the buckets of peppers Peter harvested and sees minimal visible damage. He sends the peppers to the local market and makes a profit from them.