Prismatic axis, differential-drive robotic kiwifruit harvester for reduced cycle time

Abstract

The New Zealand kiwifruit industry is currently worth over NZD 2 billion in annual sales revenue, this is forecast to double within the next 5-10 years due to a significant increase in production volume. The industry is already experiencing a prevalent shortage of labour which is indicative of a global trend. The objective for this study was to define important metrics for the robotic harvesting of kiwifruit and to propose a hardware configuration which could improve on the current kiwifruit harvesting module (KHM) developed by the Multipurpose Orchard Robotics (MOR) project. The research scope was focused to establishing whether a prismatic axis kinematic structure was more effective than a rotational axis kinematic structure for the multiple-robot harvesting of kiwifruit. KPI's (key performance indicators) were defined as evaluation and design measures - these included fruit damage, harvestability and nominal harvest cycle time. An equation specific to the KHM was derived for the latter which included measures of fruit per harvesting phase, time between fruit and a proposed work distribution constant W_D. A prismatic axis, linear rail constrained, kiwifruit harvesting robot (LHR) with two robot-arms was developed, built and tested. It was found to be exponentially beneficial to locate mass proximal to the X axis carriage centres which is achieved with a differential-drive of the YZ axis'. The prismatic axis kinematic structure of the LHR allowed for an 88% greater work distribution constant W_D, a 40% greater harvestable taskspace volume V_h and 2.5 times greater overall workspace efficiency when compared to the KHM. The nominal harvest cycle time was identical for both of these systems. However, the LHR and the developed `x-rank' registry algorithm were capable of maintaining W_D value despite a two-fold increase in robot-arm density. Therefore, in non-collision scenarios the LHR can operate with four robot-arms without compromising performance. In this scenario where both systems have four robot-arms, the LHR had a 44% reduction in harvest cycle time. Further study would need to be done into manipulability measures, scheduling methods and the effects on work distribution to establish whether a prismatic axis structure remains favourable if an orientation structure is implemented for end-effector dexterity.