Apple picking is very labor intensive and expensive. Between 15-18 billion apples are harvested every year in Washington state alone for fresh market consumption, but often farmers can’t find enough people to pick the fruit. Furthermore, laborers are hard to acquire as apple picking is seasonal. This presents an urgent need for improved harvesting techniques. Apples require a system that is delicate enough to pick the fruit without bruising it, while also maneuvering around tree branches, leaves and other obstructions. Apple-picking robots can potentially save the apple industry millions of dollars.
Engineers and scientists at Washington State University Tri-Cities and the WSU Center for Precision and Automatic Agricultural Systems (CPAAS) have invented an adaptive robot that will pick apples as efficiently as people. The robot features an arm and “hand” in which motors operate along with a vision system to delicately grasp and twist the fruit off the tree. Underactuated hands are much simpler devices that can perform a human-like grasp compliant to the object without requiring actuation of each joint. The end–effector is an underactuated, tendon–driven device designed to produce a spherical power grasp providing form–closure of the fruit. Underactuation is incorporated to provide a grasp adaptive to apples with asymmetrical geometries and variable orientations. An additional advantage of a flexure for harvesting applications is that its out–of–plane compliance helps minimize damage during unintended collisions. Passive compliance is incorporated to increase grasp robustness to error acquired during perception of the target object’s position.
The robot’s vision system incorporates cameras and sensors to capture an image of the tree. Using algorithms to identify color, shape and texture, it differentiates fruit from the rest of the plant and determines fruit location so the robotic arm can be directed for picking.
Applications and Advantages
The proposed end-effector is a relatively simple, light-weight, low-cost device that utilizes tendons to provide an adaptive grasp replicating the motion of manual apple-picking. Future work will include force analyses examining tendon tension and spring parameters, finite element simulations, and sensor selection and placement. Results from planned experimental tests to determine the dynamic forces involved during manual apple picking will be used to develop a control scheme that ensures applied forces will not damage the harvested fruit. After completing fabrication, the end-effector will be optimized in preparation for full-scale laboratory and field tests with the complete system.