Evaluated Mechanical Harvesting Systems
-
Robotics
Problem Statement
Robotic solutions for fresh market fruit and vegetable harvesting have been studied by numerous researchers around the world during the past five decades (1-5, 9-14). However, very few developments have become adopted and put into practice. The reasons for this lack of success are due to technical, economic, horticultural and producer acceptance issues. According to Sarig (5), “The major problems that must be solved with a robotic picking system include recognizing and locating the fruit, and detaching it according to prescribed criteria, without damaging either the fruit or the tree. In addition, the robotic system needs to be economically sound to warrant its use as an alternative method to hand picking.” If the plant growth systems can be modified to improve harvestability, the robotic system will have a much better chance of being successful. In order to have a successful automated/mechanized/robotic system, the cultural practices must be designed for the machine and the variety (6). Past robotic harvesting efforts in the 1980s and 1990s failed to achieve harvesting efficiencies greater than 75 to 80% (4-5, 13-14). The majority of these inefficiencies can been attributed to interior fruit which are partially or totally occluded by leaf and limb obstructions, which reduces detection efficiency, obstructions which hinder reaching target, and gripper failure, all of which reduce harvesting efficiency (5).
Background
The University of Florida, often working with the Florida Department of Citrus, has been a world leader in citrus harvesting research conducting numerous studies on mechanical harvesting and related topics over the last seven decades (7-8). In addition, UF has conducted research on automated robotic citrus harvesting over the past five decades (7,9-13). There has been a wealth of information learned during that time about various approaches to mass and selective robotic harvesting. As mentioned by Davis et al. in 1969 (6), designing the tree architecture for optimal harvesting is critical to successful economics, cultural practices and harvesting efficiencies (15). These thoughts were more fully developed in the 1970s through 1990s by various researchers among several commodities who recognized the potential that existed for designing groves and machinery for high density tree crops whose canopy size was more suitable for automation. (28-31).
The University of Florida has been conducting research in citrus grove architectures and tree factors for decades in traditional groves, as well as more modern concepts like the promising higher density production systems (15-27). These high density trees have been explored for modern concepts like the Advanced Citrus Production Systems (ACPS) and Citrus Under Protective Screen (CUPS). In addition, abscission chemical research was expected to play a vital role in harvesting success (tbd) by reducing harvesting forces which can lead to excessive tree damage thus affecting grower acceptance. Unfortunately, these chemicals which were proven scientifically successful in improving harvesting efficiency were not approved by EPA and the program was terminated. Castle and Burks of the University of Florida planted a model ACPS grove for robotic and mass harvesting research at its Plant Science Research and Extension Unit in Citra, Florida in March 2007 and then replanted in 2016. It consisted of commercial mandarin hybrid, navel, sweet orange and grapefruit cultivars on size controlling rootstocks. Scion-rootstock combinations were selected to provide an opportunity to plant at densities higher than normal ranges from 218 to nearly 300 trees/acre. The trees were managed through hedging to keep a profile suitable for robotic and mass harvesting.
The primary investigators in the early years of mass harvesting research on traditional groves were William Kahl and Glenn Coppok in the 1950s -1960s, Jodie Whitney in the 1980s till early 2000s and Reza Eshani in the 2000s to mid-2010s. In the 2010s till present, Tom Burks has explored research into Over the Top Mass Harvesting (OTMH) of high-density citrus (ACPS). These topics will be covered in more detail in Mass Harvesting. However, the research outcomes identified through mass harvesting were informative for selective robotic harvesting which is the subject of this overview.
Robotic Harvesting and Navigation Research
Dr. RC Harrell at the University of Florida in Gainesville was truly one of the world pioneers in robotic tree fruit harvesting with his work in robotic citrus harvesting in the 1980s. His pioneering work along with research that would follow in Spain and France in the 1990s laid the foundation for all robotic tree fruit harvesting that has followed (13-14). Dr. Harrell was the first to use Color Vision in Robotic Fruit Harvesting to enhance digital color images used to control a robotic manipulator for fruit harvest (9). He also developed algorithms for discriminating fruit for robotic harvest using color in natural outdoor scene which specifically investigated the use of chrominance and intensity information from natural outdoor scenes (10). Harrell also developed a robotic end-effector for picking citrus fruit (11). Meanwhile, he also developed a mobile grove-lab to study the use of various robotic technologies for picking oranges, where he used a novel sensor system and state network programming (12).
In 2002, the Florida Department of Citrus initiated a research program with the University of Florida to address fundamental technology barriers to robotic citrus harvesting. Dr. Burks was the lead PI on that effort. Working in collaboration with Dr. Castle, an FDOC funded project was initiated to develop a model orchards at PSREU in Citra, Florida and another at SWFREC in Immokalee, Florida, with a third originally planned for IRREC in Ft. Pierce, Florida. The purpose was to develop ACPS blocks in three distinct growing regions of Florida that could be used for cultivar studies as well as mechanical and robotic harvesting R&D. Eventually, only two of the sites were established due to funding limitations. One at PSREU and the other at SWFREC. Both of these sites would eventually be used for harvesting studies using Dr. Burks Over-The-Top harvesting concepts for mass and robotic approaches.
Robotics Research Barriers
Tom Burks of UF/IFAS researched several of the robotics research barriers.
Fruit Detection System Development sought to improve the ability to locate fruit in the tree canopy, overcoming some of the difficulties associated with traditional machine vision approaches attempted in the past. Bulanon et al. (32) showed that fruit visibility in the canopy can be increased by acquiring six or more different perspective views of the canopy, similar to the way human pickers locate fruits, achieving above 90% detection. Image processing techniques to separate fruits that are in cluster and to detect partially occluded fruits were also developed using visible, NIR and thermal imaging techniques with sensor fusion (33-35).
Harvesting Arm Visual Servo Control and Sensor Interface Development A harvesting manipulator has undergone preliminary field testing, where the current sensors and manipulator controls have operated under actual harvesting conditions. Functional linear and non-linear visual servo control and a statistically based 3D mapping algorithm have been developed to implement citrus fruit harvesting, path planning and execution. (38-41).
Design and Optimization of Manipulator Arm Burks used computer aided design and robotic analysis tools to select an appropriate configuration for an agricultural robotic manipulator for citrus (36-37). This forms the basis of development for an optimal citrus harvesting manipulator through advanced workspace and dexterity modeling. They fabricated and tested a hydraulically actuated macro positioning manipulator, which can be integrated in a future R&D effort.
End-effector development They conducted citrus fruit material properties, tree characteristics and conducted harvesting mechanics tests, which evaluated fruit detachment forces, peel damage criteria, and harvesting mechanics to reduce peel damage. These tests provided valuable information that were used in developing an efficient three fingered gripper end-effector for harvesting citrus, which had machine vision and optical proximity detection incorporated in the palm of the hand (36-37). In addition, they explored other end-effector designs that have likewise been field tested.
Autonomous Guidance Development The sensors (vision, Lidar, IMU, GPS and Ultrasonic) along with navigation control software were developed and tested both on a tractor platform and on a utility vehicle in the citrus grove successfully demonstrating field navigation (43-44).
References
-
Mass Harvester: Trash Removal System Development
Mass Harvesters
A pull behind platform with trash removal ability and fruit cleaning system was designed and built for mechanical harvesters. The overall system consisted of a pinch cleaner, blower, and yield monitors that aids in cleaning and removal of trash, and yield estimation during the mechanical harvesting process. This system had the ability to clean the entire fruit load from a hauling “goat truck” and completely remove all non-fruit trash. In addition, the system was also capable of measuring the amount of trash and yield (based on mass-flow). In addition to these components, an extended de-stemmer was also designed and fabricated to be used with mechanical harvesters. Preliminary studies showed that the extended de-stemmer has a removal efficiency of 50% for petioles less than 2 inches and 86% for petioles longer than 2 inches.
Key Findings:
- The recent citrus black spot (CBS) disease outbreak has required that trucks carrying fruit be tarped. Since the new pull-behind trash removal machine can completely remove leaves from the load of fruit, tarping is not needed. Tarping is used to prevent CBS-infected leaves from spreading to other groves during transit, and is an expensive and potentially hazardous process.
- The pull behind platform can aid in cleaning the trash and fruits associated with mechanical harvesting, and reduce processing costs.
-
Mass Harvester: Canopy Measurement and Control System
Tractor drawn canopy shakers can cause tree injuries as the tines strike the tree. A low-cost tine movement controller was developed to help reduce these tree injuries caused when the tines strike the tree limbs. The goal is to reduce the time required for adjusting the amplitude of the shaker (for tine control) and further automating the process. The controller controls the tines based on the distance measurement.
Key Findings
- The controller for tine distance and movement adjustment can help in rapid control and operation of mechanical harvester, especially the shakers, reduce tree injuries, and ease the operator job.
-
Mass Harvester: Analytical Model - Determine Optimal Shaking Frequency of the Tree Canopy
The simulation program for evaluating shaking characteristics can assist in developing better shaking and fruit removal mechanism.
Key Findings
- An analytical model with a simulation framework based on finite element methods was developed and validated with field-experiments.
- The results indicated that the analytical model could predict ~60% variations in acceleration values experienced by the tree.
- The model could determine the accelerations in the tree structure for different boundary conditions (machine parameters).
-
Mass Harvester: Catch Frame/Recovery Rate Improvements
During citrus harvesting, two canopy shake and catch harvesting machines work in pairs, one on each side of the row of citrus trees necessitating synchronization of their travel speed and steering. Unreliable synchronization causes inefficiency in the catching system. This degrades the capability of the citrus harvesting system. To retrieve the fallen fruit, additional labor and time is required. This is one issue that lowers the canopy shaker’s harvesting efficiency. In recent years, mass harvesting fruit removal efficiency has been reported up to 95%, while the catch frame recovery efficiency has been reported between 88% and 92%. In recent grower observations, noted by Roka, catch frame efficiencies have been dropping into the low 80%, and in some cases even below 80%. The potential reasons for catch frame losses are numerous.
- Misalignment of forward/reverse synchronization of left and right vehicles
- In/out gap between catch frame seal and tree trunks on either side
- Vertical misalignment of catch frame seals due to grove terrain, especially considering the bed-top and swell-bottom elevation differences.
- The effectiveness of the fish-scale system employed to seal tree trunk
Key Findings
The specific accomplishments were the following:
- The design concepts developed in 2008/09 have been further modeled and developed for a new catch frame closure system and the frame position control.
- A scaled prototype of the new catch frame system is being developed to demonstrate the concept. The concept consists of a new closure material which should be more compliant to the tree trunk, and a three section micro-adjustment which improve closures around the tree, while still protecting resets.
-
Mass Harvester: Canker Decontamination
The overall goal of this work is to study the means and ways of facilitating and expediting the canker decontamination process for mechanical harvesting systems. The immediate goal is to study the citrus mechanical harvesters and to identify the areas on the mechanical harvesters that tend to collect plant materials and then work with the manufacturers to provide improved designs and adjustments that can reduce the plant material collection on the mechanical harvesting systems.
The tractor drawn continuous canopy shaker, manufactured by OXBO International Corp, Clear Lake, Wisc., was studied by recognizing the areas of the machine where most plant materials are collected and suggestions were provided to the manufacturer. The manufacturer, OXBO, acted on the suggestions and provided covers in order to restrict the collection of plant materials by the harvesting machine. There was a substantial decrease from an average of 8.5 pounds to 2.5 pounds of plant materials. Engineers and scientists of UF CREC also undertook similar studies for canopy shake and catch from the same company.
The Coe-Collier trunk shake and catch harvesting machine was studied and suggestions were made. After the covering was placed on the frame, the collection of plant materials substantially decreased as shown in the figures below. Further modifications were explored on this machine. Meanwhile, growers using mechanical harvesting aids were encouraged to undertake similar studies on their machines.
Key Findings
- Substantial reduction was realized by adding covers to harvesters dropping from 5 pounds to 2.5 pounds of plant materials.
-
Mass Harvester: Multiple Shift Harvesting
The key objectives of this topic were to:
- Evaluation of issues and challenges to the adoption of multiple shift (24 hr/d) mechanical harvesting.
- Exploration of the feasibility of developing a stochastic equipment utilization model to optimize producer/harvesting/hauling/processing aspects of the harvesting chain operation, and thus achieve the most favorable economic return on investment.
- Development of sensing and virtual reality technologies capable of enhancing equipment utilization, operational safety, and tracking during 24 hr/day harvesting.
- An issues paper associated with 24 h/d Mechanical Harvesting was planned
A white paper was to be generated, which would document the issues associated with implementing 24 hr/day citrus harvesting. Discussions with the citrus industry would provide valuable insights into the challenges that will be faced by growers, processors, and harvesting managers when attempting 24 hr/day harvesting. Additional discussions was to be conducted with members of the sugarcane industry who have adopted 24 hr/day harvesting to better understand the logistics and safety issues that have been encountered in that industry. The outcome of this paper will help further define the goals of this program. As a part of this work, preliminary discussions are occurring to determine the need for a stochastic modeling environment for resource utilization.
A significant challenge associated with nighttime harvesting for both logistics planners, and operators in the field is keeping track of other equipment, monitoring field conditions, and avoiding dangerous obstacles, which can be easily seen in the day. It was proposed to begin work on a virtual environment, which will help operators stay alert for changing field conditions, and logistics planners track equipment in the field. The system will rely on devices such as machine vision, laser radar, RFID, DGPS, and wire-less technologies to spatially locate equipment in the field. A virtual environment can then be used to monitor equipment locations, and dangerous field conditions such as irrigation ditches. This technology will be especially beneficial to goat drivers who must transport harvested fruit from the MH system to roadside. During the first year they intended to develop a fundamental virtual environment, and initiate exploration of technologies required for creating a grove virtual environment.
Key Findings
- A series of interviews were conducted with growers, harvesting contractors, trucking companies, juice processors and packinghouse managers to access the impact that multiple shift mechanical harvesting would have on the industry. Once the interview process was completed, an issues paper was planned.
- Preliminary concepts were formulated for the development of a stochastic modeling environment for optimizing resource utilization in the citrus harvesting process. This work was to be done in conjunction with an economics model.
- Sensing technology and computer software development was planned to enable a scouting-based virtual reality representation of the grove environment, which may be useful for harvesting efficiency, safety, machine tracking, and utilization modeling reasons. This technology should have applications for mechanical and robotic harvesting, as well as for disease and pest scouting
-
Mass Harvester: Pick Up Machines
The key objectives of this topic were to:
- To evaluate a ground fruit pick up system developed by Oxbo International Corp. The evaluation was based on the performance of picking up the desired fruit and its efficiency for removing undesirable fruit and trash under different ranges of forward speed, orange variety, and grove conditions.
- To evaluate the external microbial load of ground fruit picked up by this machine.
Various mechanical harvesters and pick-up machines have been developed since 1970. The tractor drawn trailed canopy shaker detaches and allows the fruit to fall on the ground. Trunk shakers and blowers are also used to detach the fruit. These harvesters are used in old, non-uniform trees with or without skirting. The fruit is then picked up by a hand-crew or pick-up machines. Due to a large volume of fruit on the ground, a pick-up machine is preferred for the job. Its importance was also felt when large numbers of fruits dropped on the ground from trees by hurricanes.
Some of the rake and pick-up machines were developed in 1970-80. These machines can pick up more than 400 lb/min with approximately 92% picking efficiency. The Department of Citrus in Florida has developed a rake and pick-up system which was acquired by CREC, UF/IFAS. The pick-up machine was tested recently and performed well. The challenges faced by above mentioned systems are the width (to 20 ft) and size which restricts them to operate in modern groves. Moreover 2 to 3 persons are required to operate the system.
OXBO International Corp. (Oxbo), Clear Lake, Wisc., had developed a citrus fruit pick-up machine. It was a self propelled rake, pick-up and cleaning system with a suitable size to operate in modern groves. The company and engineers from CREC are studying its operational details and will soon be ready for field evaluation by CREC. Evaluation of these machines in actual field conditions for a prolonged period would ensure confidence in the growers to adapt these machines as it involves a great amount of initial investment.
Key Findings:
- Study results were in progress at time of this posting.
-
Photo Gallery