Dan Wheat/Capital Press
Yet another innovator seeking to develop a robotic apple picker doubts three others will have commercial machines soon and believes his partially-robotic concept is more readily attainable.
Minglei Xu, 42, was born and raised in Shanghai, China, and obtained his master’s degree in computer science from Brigham Young University, Provo, Utah, in 2000. He worked in Silicon Valley before becoming a serial entrepreneur in 2010. He built the gaming studio Wild Needle and sold it to Zynga in 2012. He started Grow Mobile — an ad buying platform for marketers to buy, track and optimize mobile app promotions — in 2012 and sold it in 2014.
The next year, Xu settled on a robotic apple picker as his next challenge because of the need given labor shortages and because apples are relatively common in shape and size versus dissimilar objects that are more difficult for robots to pick. He started the company FruitBot in Palo Alto, Calif.
“Picking apples is very repetitive and grueling work that the robot is very good at,” Xu told Capital Press.
More resources are being deployed into the navigational side of Artificial Intelligence such as self-driving cars, but manipulative AI such as robots capable of cooking or picking fruit without rigid coordinates is the key unsolved arena, Xu said.
Abundant Robotics of Hayward, Calif.; HarvestMoore of Pasco, Wash.; and FFRobotics in Israel all hope to have robotic apple pickers ready for commercial use in late 2018 or early 2019.
They may be able to reach that goal under very constrained conditions but won’t be able to achieve the picking rate they want, Xu said.
“A fully automated robotic picking system in any industry is nowhere close. It’s five years away at least,” he said.
A big problem is avoiding obstacles such as limbs and leaves that a robot mistakes as fruit, Xu said. It is seen as a two-dimensional issue that requires aggressive pruning and thinning to remove obstacles when it’s really a three-dimensional problem, he said.
While others are targeting the picking of apples within a depth of 18 inches inside the canopy, he said he is working toward a 36-inch depth. His solution is a robot or two robots on the front of a harvest platform assisted by several human pickers behind. Robots pick the low-hanging, easy fruit and humans pick what the robot misses.
Such a system could be commercialized in 2019 with a 60 to 70 percent pick rate, he said. Robots are not so good at apple clusters and could beep to alert humans to pick those, he said.
“The robot speeds up the most repetitive and easiest part of the work,” Xu said.
It’s a transitional move that doesn’t require a big a leap of faith by growers as a purely robotic system does, he said. Mike Willett, manager of the Washington Tree Fruit Research Commission in Wenatchee, said he wrote in a letter in support of Xu getting a Specialty Crop Research Initiative grant.
“A lot of people are looking at pieces of the puzzle. There are a lot of issues. Harvest is one,” Willett said. “There’s thinning, blossom thinning and different tools at the end of a robotic arm. To move it all forward, we need to have lots of smart people thinking about the problems.”
Xu wants to focus solely on software and robotic picking and says conveyance of fruit into bins, field sorting and stem clipping may include humans and can be figured out by companies already engaged in farm equipment such as PIUMA in Italy and Automated Ag Systems builder of the Bandit Xpress harvest-assist platform in Moses Lake, Wash.
Xu said he has spoken with both but has not started working with either.
He foresees developing systems to not only help in harvest but logistics and big data, that is tracking of each piece of fruit to provide much more information to marketers and retailers.
Xu is working with five growers in Washington and California for field trials but wants to visit more orchards to look at differences in tree architecture and fruit varieties.