Apple Picking Using Dual Robot Arms Operating Within an Unknown Tree

Abstract. Agriculture has begun to rely more heavily on robotics for harvesting. Apple harvesting poses a particular challenge in fruit acquisition, as the branches of the tree are hard to map and navigate. This research focuses on creating a dual arm robotic system that can successfully navigate tree branches to quickly yield more apples than what is visible from outside the canopy. The experiments performed were done in the computer simulation environment V-REP, and a new form of path planning was developed specifically designed for the challenges faced in apple harvesting. The only external sensors are RGB cameras, located on the end-effector of two cooperative manipulators, and it is assumed that the cameras are sufficiently high resolution to locate the position of the apples in pixel space. With this information, a navigable network of clear paths from within the branches is developed for a manipulator to reach the fruit. This method, which this research refers to as the “Void Space” path planning algorithm, was performed for 50 trials of 20 apples randomly placed among the branches. The experiments performed demonstrate that while only an average of 68% of the apples were visible from outside the canopy, a manipulator could be made to reach approximately 92% of the total apples generated. With a properly fast and dexterous manipulator, the use of Void Space path planning could greatly increase the harvest yielded with minimum impact on the time to reach each individual fruit.