Loading...
An efficient process to designing robotic end effectors for high value crops: Application on robotic apple fruitlet thinning
Abstract
High value crops continue to rely extensively on manual labour for labour intensive and expensive tasks like harvesting and thinning. Robotics is increasingly being investigated as a solution to combat labour shortages which threaten the sustainability and growth of the horticultural sector. However, the seasonal nature of high value crops offers only a short window of a few weeks to field tests one robotic end effector per season using the conventional process. This hinders the research and development timeline of robotic end effectors. This thesis presents an efficient design process, modified from existing design methods, to overcome this limitation by capitalising the remainder of the off-season window within a year, and thus reduces the robotic end effector development time.
The efficient design process starts with co-design workshops involving growers and field visits to gather requirements, understand crop physiology, and determine the specific needs for identified crop management tasks. This information guides the development of several end effector concepts. Concepts with feasible potential are selected to progress to the prototype stage, which then undergo laboratory testing and comparative evaluations during off-season. The off-season investigation can only be possible by creation of an artificial crop structure that replicates the crop's physiology and is capable of testing the prototype's core mechanism principles repetitively. This structure is also optimised with enhancements and design optimisation of the end effectors, such that field testing time can be maximised to address issue unique to field conditions. This process then allows for extensive testing of various end effectors within a single year, significantly reducing the development duration. The process prioritises the generation of multiple viable end effector concepts while integrating feedback from growers.
The efficient design process was implemented to develop robotic end effectors for apple fruitlet thinning. The design requirements for thinning fruitlets to 1 or 2 fruitlets from a cluster were identified after consultation with growers and a visit to the apple orchard. Nine concepts were generated and preliminarily tested in the field and consulted with growers. Four distinct concepts were then selected: 1) cutting that cuts the stalk, 2) suction based on grasping and rotating, 3) paraboloid based on grasping and rolling, and 4) piercing end effector based on spear piercing and rotate. These concepts were further developed into fully functional end effectors for laboratory evaluation, enabled by the development of an artificial structure that allows interchangeable fruitlets of all sizes and accommodating repetitive testing of the end effectors core mechanism principles. After an iterative process of testing, optimisation, and modifications of end effectors and artificial fruitlet structures, a comparative evaluation was performed to benchmark and quantify the capability of each end effector was conducted.
A computer vision system and path planning systems were developed and integrated with the end effectors to facilitate this evaluation. This shows that the vacuum suction end effector consistently reached a 100\% success rate from 90° pitch angle on single fruitlets, and also maintained this success rate at a 90° pitch angle in a fruitlet cluster in all positions, followed by cutting and paraboloid with 80\% success rate on a single fruitlet. In contrast, the spear piercing mechanism consistently under performed and was excluded from the field testing. The field test in apple orchards under real world conditions indicated that the suction end effector achieved the highest success rate, reaching 70\%, followed by cutting end effector with 44\% and paraboloid with 26\%, showing trends consistent with lab evaluations. The primary causes of failure are positioning errors and occlusion. Thus, the efficient design process developed the suitable suction end effector to be integrated with the overall robotic platform for further development within a single year, compared to a minimum of two years using a conventional process.
Type
Thesis
Type of thesis
Series
Citation
Date
2025
Publisher
The University of Waikato
Supervisors
Rights
All items in Research Commons are provided for private study and research purposes and are protected by copyright with all rights reserved unless otherwise indicated.