Work distribution of multiple Cartesian robot arms for kiwifruit harvesting

Abstract

Employing multiple robot arms for kiwifruit harvesting can raise the efficiency since the task completion time is shortened. However, task partitioning and reachability are two major concerns. It is found that the task completion time is minimized if task partitions for robot arms are uniform. However, the partition uniformity is influenced by the fruit indivisibility and fruit cluster growing style. It is also constrained by the fruit distribution across the canopy and the robot arm positions which affect the fruit reachability. This article investigates how to partition the tasks of kiwifruit harvesting so that the harvesting can be completed by multiple robot arms in the minimum time with an assumption that the fruits can be harvested by any robot arm. A parameter termed work distribution is introduced to measure how the task partitioning is different from the optimum. A research platform with two robot arms is implemented to show how to approximate the assumption. Ten field experiments of kiwifruit harvesting had been run by the research platform in two kiwifruit orchards in New Zealand with satisfactory outcomes.