—by Ross Courtney
Every year, apple fruit thinning overlaps with cherry harvest, posing tough labor management decisions.
Northwest engineering researchers are working on a bot for that.
Students and researchers at Oregon State University and Washington State University are developing a robotic fruitlet thinner with a computer-vision-guided robotic arm that identifies fruitlets and reaches out with a claw-like end effector to yank them off the tree and drop them to the ground.
In fact, the engineering teams are working on a robot that would take care of several apple functions — pruning, cluster-by-cluster pollination, blossom thinning and fruitlet thinning. All use a similar arm with six points of articulated motion guided by computer vision and machine learning.
They started off as individual projects with different funding sources but have been combined in recent years into the multipurpose orchard robotic system, said Manoj Karkee, director of WSU’s Center for Precision and Automated Agricultural Systems in Prosser, called CPAAS for short.
These management tools could also be incorporated into harvest robots by swapping out the end effectors. A multipurpose robot approach will make it easier for growers to eventually justify the cost of an expensive machine, Karkee said.
The pollinator robot has an end effector that identifies flower clusters and sprays them with pollen; the blossom thinner does the opposite with a small brush that breaks up the clusters. The pruner senses tree branches and … prunes.
The researchers are still in the early stages, but eventually they hope to partner with a private company that can incorporate the technology into commercial machines. Karkee’s team has experimented with putting their end effectors on one of the harvest machines in commercial development.
Karkee estimates his lab has spent about $3 million over the past 10 years with grants from the Washington Tree Fruit Research Commission and U.S. Department of Agriculture’s National Institute of Food and Agriculture.
The fruitlet thinner
The robotic thinner has an articulated arm that swivels and bends to reach out for the identified fruitlet and pull it loose from the tree with a 3D-printed plastic claw that has two fingers. Video camera sensors with machine learning identify the apples with a probability of certainty and direct the arm.
The machine is years from commercial application.
During an early June demonstration for Good Fruit Grower, graduate students meticulously lined it up and debugged the software before every pick. The claw sometimes missed its target apple and once removed a spur.
Karkee puts it at about three on the readiness level, a 1–10 measurement of technological maturity. The robotic pollinator, blossom thinner and pruner are a little ahead of that, somewhere between three and four.
But perfection is not really the goal. The eventual machine just has to be good enough for a reasonable return on investment. A grower might find that if the robot successfully handles, say, 90 percent of the task. That threshold drops as more functions are added to a machine.
OSU involvement
At OSU, researchers bring their expertise in developing digital and physical “proxy” environments to train robots year-round.
Relying only on real-world experiments limits data collection and training to only a few weeks or months per year, said Joe Davidson, assistant professor of robotics and head of the Intelligent Machines and Materials Lab in Corvallis.
Davidson and Cindy Grimm, a robotics professor, have built fake trees made from springs and magnets to teach robotic end effectors to mimic the picking motion of a human hand on real apples. Their teams also built virtual orchards to train computer algorithms on the “perception pipeline,” how a robot recognizes and understands tree architecture from sensor images.
They plan to field-test those algorithms later this year, Davidson said.
They have spent roughly an additional $1.5 million, also from the research commission and USDA-NIFA, over the past six years on the related efforts, Davidson said. •
Leave A Comment