Mechanical Design and Field Evaluation of a Robotic Apple Harvester

Joseph Davidson

Every apple destined for the fresh market is picked by the human hand. Despite substantial research to develop robotic apple harvesters, there are no robotic systems commercially available. The highly unstructured orchard environment has been a major challenge to the development of commercially viable robotic harvesting systems. The absence of mechanical harvesters is a significant concern due to rising production costs and increasing uncertainty about the future availability of manual labor. This dissertation presents the mechanical design of a robotic apple harvester. The overall approach implemented was to advance performance, as measured by speed and harvesting efficiencies, by simplifying the harvesting task. A custom, seven degree-of-freedom manipulator was designed, fabricated, and then integrated with a picking end-effector. The end-effector, which is the only system component that makes contact with the fruit, is an underactuated, passively compliant design that grasps the fruit with a spherical power grasp. The end-effector prototype was extensively analyzed in the lab prior to field testing and shown to be robust to perception error. Prior to field testing, a global camera was integrated with the mechanical system in order to execute open loop, go-to picking with no intermediate visual servoing. The system was then evaluated in a commercial apple orchard in Prosser, Washington. Robotic manipulation adopted ‘undersensed’ picking methods developed through dynamic analysis of the hand picking process. Because modern, planar orchard systems were selected for field studies, substantial computational resources were not dedicated to motion planning. Detailed performance criteria were used to report results, and each significant task in the harvesting process was individually timed to help focus future efforts at reducing cycle time. The system successfully picked 127 of the 150 fruit attempted for an overall success rate of 84%. The average picking time was 6.0 sec per fruit. These fruit detachment efficiencies and execution times represent an approximately 3 sec improvement in performance levels reported for robotic apple harvesters. However, substantial challenges to commercial implementation still remain. An overview of future work needed to address some of these challenges is included in the summary.

Mechanical Design and Field Evaluation of a Robotic Apple Harvester

Files and links (1)

Abstract

Metrics

Details