Qin Zhang, Manoj Karkee 2016-10-25 00:24:16
The U.S. tree fruit industry is an important component of the nation’s agricultural sector, representing about 10% to 13% ($14 to $18 billion) of all crop production. Currently, fresh market fruits are harvested manually around the world, which makes the industry highly labor intensive and less sustainable due to rising labor costs and increasing labor shortages. In the Pacific Northwest region of the U.S., labor costs for harvesting fresh market fruit account for 20% to 30% of all on-farm variable costs. Researchers in the past have investigated different technologies for mechanical harvesting of tree fruit. Despite wide differences in the specific mechanisms, almost all the investigated systems were designed to perform either shake-and-catch harvesting (mass harvesting) or pick-and-place harvesting (robotic harvesting). In principle, shake-and-catch harvesting applies vibratory excitation to the canopy, trunk, or a branch of a tree to create a detaching force on the fruit and uses some type of catching device to collect the detached fruit. Mechanical tree fruit harvesters based on this approach are used to harvest fruit destined for the processing market due to the high productivity of shake-and-catch harvesting. However, there has been limited success in fresh market fruit harvesting due to the high level of harvest-induced damage. For pick-and-place harvesting, the process includes locating target fruit, approaching and detaching the fruit, and then placing the picked fruit in a designated container. In general, robotic harvesting has achieved very limited success, primarily due to inadequate accuracy, speed, and robustness. Because none of the existing technologies has provided a satisfactory solution, mechanical harvesting of fresh market fruit remains a crucial problem for the long-term sustainability of the tree fruit industry. To develop a roadmap for fully automated mechanical harvesting systems that would be practical for harvesting fresh market fruit, a joint task force consisting of an industry advisory group and a technology development group from Washington State University has specified that an ideal solution should be capable of harvesting more than 95% of the fruit with less than 5% harvest-induced cullage in SNAP (Simple, Narrow, Accessible, and Productive) fruiting wall orchard systems, using less than 20% of the current level of human labor. In addition, a system capable of achieving a harvest speed of one fruit per second should be economically competitive with current harvest methods. While fully autonomous harvesters navigating between tree rows are still years away, researchers at Washington State University’s Center for Precision and Automated Agricultural Systems (CPAAS) are working on the core technologies for fully automated mechanical harvesting systems, both for mass harvesting and robotic harvesting. Shake-and-catch harvesting Fully automated mass harvesting, using a shake-and-catch system that shakes only a targeted canopy region and catches the fruit directly beneath, could achieve high productivity at a relatively low cost if the harvest-induced fruit damage could be controlled within an acceptable tolerance, which is a crucial challenge for this technique. Developing this type of harvesting system with minimal fruit damage requires fundamental understanding of (1) the achievable productivity and fruit quality from a shake-and-catch process, (2) how to control the system parameters to achieve such a result, and (3) tree architectures that could optimize the horticultural attributes and be machine-friendly for automated mass harvesting. Given the multi-faceted nature of the problem, the CPAAS team used a trans-disciplinary approach to gain a comprehensive understanding of the physical and biological aspects, and then used that knowledge to create possible solutions for the recognized challenges. One of the important findings was that an appropriate combination of shaking pattern and rhythm could improve localized fruit removal from the targeted tree branch and also reduce harvest-induced fruit bruising. Field testing in a formally trained Gala apple orchard revealed that semi-selective shaking could achieve 88% fruit removal with less than 4% shaking-induced bruising. Another test using a prototype fruit capturing device on a vertically trained Jazz apple orchard showed that it was possible to remove 92% of the fruit from the targeted canopy region, collect up to 99% of the removed apples under the shaken branch, and maintain 85% of the collected apples at the highest extra-fancy grade. By keeping the harvest-induced fruit damage rate to less than 15% for several varieties and about 5% for a few varieties, this research indicated the possibility of achieving the desired fruit quality with mass harvesting using appropriately designed fruit capturing devices. Pick-and-place harvesting Using a three-step process of locating, detaching, and placing of fruit, robotic pick-and-place harvesting offers the possibility of selectively harvesting individual fruit. This technology has been extensively studied since the 1980s, with numerous prototypes evaluated. However, very few systems are ready for commercial adoption, largely due to unsatisfactory productivity, robustness, and fruit quality. While all three steps affect the results, the technologies for fruit locating and placing have been well studied and validated, at least at a conceptual level. The middle step—fruit detachment—is the most challenging task and remains the central challenge in robotic harvesting. Most studies have used a modified industrial robot to detach fruit, which could be one of the reasons for the resulting performance deficiencies. Fruit picking may require a more complicated grabbing and detaching motion than other industrial picking applications. Based on previously reported success, a robotic apple picker being developed by a CPAAS team has achieved a cycle time of six seconds to pick an apple from a formally trained fruiting wall tree canopy. However, this speed does not yet meet the desired productivity level. To understand the effect of the fruit detachment method on picking productivity and fruit quality, CPAAS researchers conducted a study of fruit picking dynamics. The study found that the required grabbing force, on average, could vary by up to 100% for different grabbing methods while picking different varieties of apples, which resulted in picking-induced fruit bruising rates that varied from 0% to 60%. However, the grabbing patterns that required less force generally required a detaching motion that was more complicated than simply pulling the fruit off the branch. Other alternatives To make fully automated harvesting systems practical for fresh market fruit, a more creative approach is required, rather than simply adapting technologies from other applications. One area for innovation could be an overall system integration approach, including human-machine-tree interaction and improvement in horticultural systems. If we can create a technology that allows human operators to cooperate with robotic machines, then the human operator could simplify the tasks that the robotic machine would otherwise need to handle autonomously, which in turn could make the robotic harvesting solution more adaptable and affordable. For example, work is being done at CPAAS with machine vision systems that can detect and locate most of the fruit, but human operators can step in to identify the last 1% to 2% of the fruit that are obscured in the canopy and difficult for the machine to find. In recent years, many companies and venture capitalists have been attracted to developing robotic solutions for fresh market fruit harvesting. For example, Abundant Robotics, Inc., a California-based company, is developing a robotic apple picker with support from the Washington Tree Fruit Research Commission. This project involves a vacuum-based picker that can avoid damaging the target fruit, adjacent fruit, or part of the tree—which a grasping machine might do—and it achieved a picking speed of one fruit per second. Other companies around the world, such as FFRobotics in Israel, are also putting development effort into robotic fruit harvesters. Based on the recent progress in agricultural technology, fully automated tree fruit harvesting systems could become commercially available for U.S. growers in the next three to five years. ASABE Fellow Qin Zhang, Professor and Senior Scientist, and ASABE member Manoj Karkee, Associate Professor, Center for Precision and Automated Agriculture, Department of Biological Systems Engineering, Washington State University, Prosser, USA, email@example.com, firstname.lastname@example.org.
Published by ASABE. View All Articles.
This page can be found at http://bt.e-ditionsbyfry.com/article/Fully+Automated+Tree+Fruit+Harvesting/2621113/350793/article.html.