Fruit orchards are complex environments, which makes automation difficult. But two robotics projects recently awarded grants from USDA could eliminate some of the barriers to automated orchard management.
The University of Minnesota (UMN) received a grant of $914,565 to develop planning algorithms that would allow robots to operate autonomously in apple orchards.
Ibrahim Volkan Isler, an associate professor in UMN’s Computer Science and Engineering department, is leading the robotics and sensing portion of the project, along with Ai-Ping Hu, an engineer with Georgia Tech. James Luby, a UMN professor and fruit breeder – along with professors Cindy Tong and Emily Hoover – is providing a horticultural perspective.
The project started from conversations Isler had with Luby and other horticultural colleagues, about the potential of robotic applications in orchards. Research started in 2014, with help from a small seed grant from the state government. After proving the concept could work, the team applied to USDA (in connection with the federal National Robotics Initiative), which provided the funds for a three-year project late last year.
The technology the team is trying to develop will focus on two tasks: counting apples in an orchard and measuring their diameter. The end goal is to have the technology attached to a device – a camera-equipped drone, for example, or a smartphone or similar handheld item – that takes images of count and diameter, images that can be easily accessed by the grower, Isler said.
Luby said they hope to develop a system where growers can have a row-by-row, or even a tree-by-tree, estimation of yield, canopy volume and fruit size distribution at various stages of crop development, based on relatively inexpensive but comprehensive data acquisition. The data could inform important management decisions such as thinning, pruning, harvest management and marketing.
The technology’s up-close examination of fruit could be used to detect brix, dry matter content, internal insect damage and external issues. That information could inform potential estimates of pack-out, maturity and quality, Luby said.
Luby said he’s “amazed” at what is becoming possible with the sophistication and miniaturization of sensors, and that robotics equipment is becoming rapidly cheaper and better in quality, just as it has in the past for computers, cellphones, and other technology.
Field crops producers already have technological capabilities they can take advantage of, but these have not yet been developed for specialty crops, partly because of cost and partly because of the different production systems and information needed. Nevertheless, specialty crops can learn from what is happening in field crops and adapt what is useful, he said.
As a fruit breeder, Luby is excited about the project’s potential for breeding programs. There’s an “explosion” of genomics information becoming available in fruit crops, and the ability to phenotype, or measure the plants or fruits themselves for important traits, is becoming more and more critical to improving efficiency, he said.
The University of California, Davis, received a $1,069,598 grant to “develop theoretical and technological tools that will enable the design, optimization, prototyping and field-testing of consistently high- throughput, cost-effective mechanized harvesting systems for modern orchards,” according to USDA.
Stavros Vougioukas, an assistant professor and engineer with the University of California, Davis, is leading the project, along with UC’s David Slaughter and Fadi Fathallah and Stephen Nuske of Carnegie Mellon University. It’s a collaborative effort with Carnegie Mellon, Vougioukas said.
The three-year project couldn’t exist without the USDA grant, which was awarded last fall. The goals are to introduce robotic technology in orchard platforms to increase efficiency and add value to their use. The first goal will be achieved by automatically controlling each worker’s vertical positioning during picking, in order to match the incoming fruit density to individual picking speed. The second goal will be pursued by developing technology that maps fruit yields, according to Vougioukas.
The idea for the research came from California’s fruit industry. Mechanized platforms are not that widespread in the state’s orchards, one reason being that the efficiency gains don’t justify the cost of the machine. And many of the state’s orchards – processing peaches, for example – are still relatively low density, which doesn’t mesh with mechanization, he said.
One of the problems the project hopes to solve: When you have pickers on a platform, the slowest picking rate becomes a bottleneck for the entire machine. How do you alleviate that bottleneck?
Each picker on a platform has their own picking rate, which is a function of two things: the individual picker and how much fruit is in front of them. The idea is to develop an automated system that measures how much fruit a picker is picking in real time, and another system that estimates the density of the fruit approaching as the platform proceeds. A computer, by controlling the pace of the platform and the elevation of each picker, could move slower pickers to the less dense fruit areas, or combine pickers at the denser canopies, Vougioukas said.
The team’s approach is to use an existing (Bandit Xpress) platform and retrofit it with small, low-cost hydraulic lifts to control the elevation of each picker, as well as a vision system to gauge fruit density. A central processing unit would control the technology, he said.
If the team can develop a worthwhile prototype, a company could adopt the technology and turn it into a product that could be used on any commercial platform. The team is working with apples, pears and cling peaches – but the technology could apply to any orchard crop, provided the architecture is appropriate, Vougioukas said.
— Matt Milkovich, managing editor