Control, Robotics and A.I. in Agriculture
Intelligent autonomous spraying robot
-
A typical sprayer sprays a certain amount of pesticide in all directions considering the surrounding environment. As a result, various problems, such as acceleration of soil acidification and economic loss due to pesticide wastage are prevalent. To overcome these problems, we propose a deep learning-based intelligent spraying system that sprays pesticides only on fruit trees. The RGB image acquired by the camera is divided into a 4x4 area image to acquire the volume of a fruit tree in that area, and each mapped nozzle is controlled through an on/off control and variable flow control. Here, variable flow control applies only the optimum spraying amount according to the orchard environment such as the fruit and tree. Specifically, variable flow control maintains the performance of the conventional spraying method (i.e., all nozzles open) and sprays undesired pesticide to a minimum. Also, SLAM was also applied the intelligent spraying system so that the spraying can be self-driving in orchards. The proposed intelligent spraying system was validated in a pear orchard and, task allocation studies are underway to apply this system to multiple robots to increase spraying efficiency in large orchards.
Autonomous Fruit-Vegetable Harvesting Robot
-
Agricultural work requires a lot of manual labor, and therefore, research on harvesting robots to automate harvesting is active. Research on manual labor replacement with robots is usually focused on the robot’s ability for crop recognition and approaching the crop for harvesting. However, it is also important to develop an end-effector that can harvest a variety of fruits and vegetables. In this research, we propose the end-effector of a mechanism that can be stable during a harvest without additional and complex control for the various fruit and vegetables. To achieve the process of catching and cutting crops like humans, an end-effector was developed to suck fruits with suction cups and cut them with scissors or saws, in addition to research conducted on a plucking-type soft gripper that grabs and picks fruits. We designed an end-effector that includes structures that allow a stable catch for collecting falling crops after detachment from the pedicel, reducing the harvesting time and improving productivity
Semi-autonomous Korean Cabbage Harvesting Machine
-
Korean cabbage is a head-type plant, similar to cabbage. And therefore, prior research has been conducted on single-row type harvester similar to a cabbage harvester; however, it is difficult to find a commercialized model based on this research. The reason is that while the stem position of a cabbage is located outside the head, meaning that there is a low possibility of damage to the head due to cutting, Korean cabbage has a stem position inside the head. Therefore, Korean cabbage must be cut at an accurate position, as there is a higher possibility of damage to the head. However, farmland environment is highly volatile and difficult to predict. In this research, a mechanism was designed for independently-controlled cutting devices during harvest. The proposed mechanism can move the roll and pitch angle and the x-axis length; non-linear control methods, such as backstepping control, are applied to reduce the degradation of hydraulic systems due to their dynamic properties. To verify the proposed system, a field test was conducted in an actual Korean cabbage field. We harvested sixty heads in total, and as a result, high cutting performance was confirmed with 90.7 out of 100 points. Furthermore, it was confirmed that attitude control was adjusted according to the cabbage field environment. The cutting attitude was maintained within the optimal cutting position range. The proposed backstepping-based attitude control system allows accurate Korean cabbage harvesting and is expected to improve the harvest success rate.