Dr. Derek Morikawa
In developing a robotic apple-picking system, Vision Robotics Corporation has taken a revolutionary two-step approach that involves sending a robotic scout through the orchard first to map the fruit and devise a picking plan that a harvesting robot can follow.
"This idea of scouting first and knowing where everything is so you can pick efficiently and economically will be a breakthrough," Vision Robotics Chief Executive Officer Dr. Derek Morikawa told the Washington tree fruit industry a year ago.
Originally, the idea was to develop the scout as the front end of a complete harvesting robot, but it became apparent that the scout itself could be useful apart from the harvesting system.
For example, it could scan sample trees during the growing season to compile accurate crop estimates, which would be useful for both growers and packers. Packers could know the sizes and quantities of fruit in a specific orchard before the crop was delivered to the warehouse.
The scout could potentially be used to record the color of the fruit for color picking. With different types of sensors attached, it might be able to assess the internal quality of the fruit on the tree, or scout the orchard for insects and diseases.
California-based Vision Robotics has been working on the scout for two years. In the fall of 2007, the company’s scientists were in Washington State to test the scout’s stereoscopic camera system, which was mounted on a stationary boom. It was tested on several red apple varieties to prove that it could locate the apples on trees.
This fall, the Vision Robotics team was back in Washington with a modified camera system on a mobile boom to photograph trees while it was towed down the row at between a half and one mile per hour. The scientists operated it in daylight and darkness, and tested its ability to distinguish green apples (Granny Smiths) from the foliage. In addition to cameras, the system has sensors that indicate exactly where it is and whether it is tilted as it moves down the row, so that it can create an accurate three-dimensional map with the location of every piece of fruit and its size.
"This year, what we expect to do is demonstrate that we can operate the system at speed, which is a critical step," Morikawa said. Eventually, the travel speed will be around two miles per hour. It will require only a low-powered vehicle, such as an ATV, for towing.
The scout, as tested, takes 15 images per second. Morikawa said fewer pictures will be needed in orchards with fruiting walls, where the fruit is clearly visible, than in more dense canopies, where some of the fruit can only be glimpsed between the foliage.
The next step in the four-year scout project will be to have two booms working together on either side of the tree row so the cameras can see fruit on both sides simultaneously. More work will be done on the lighting to make sure the scout can detect fruit in difficult light conditions, such as when the sun’s angle is low and it is shining directly into the cameras.
During the final year of the project, Vision Robotics will focus on making the scout into a rugged, dustproof, easily maintained field product and on making sure the output of the scout is easy to retrieve and in a format that is useful to growers. Morikawa said the data would fit into mapping programs so that the grower could have a map of the orchard with a count of the fruit and its grade and size.
The Washington Tree Fruit Research Commission has provided funding of almost $650,000 for the project so far. The California Citrus Research Board is also helping to fund the scout research, as a similar device will be part of a citrus harvester that Vision Robotics is working on.
Vision Robotics is collaborating in a research project on comprehensive automation of growing practices in specialty crops that was recently awarded $6 million in federal funding. The grant, obtained through the Specialty Crop Research Initiative, will include funding to complete the scout project, Morikawa said.
The team that received the grant includes research partners who can play key roles in developing the scout system, he said. Carnegie Mellon University, the lead partner in the project, has developed an all-terrain vehicle that could tow the scout autonomously. Another partner develops wireless networks in orchards and can retrieve data from equipment such as the scout. Yet another is developing pest sensors that could be placed on the scout.
Morikawa said one of the first things his company did at the project start was to make sure the robotic harvesting system made sense economically from the grower’s perspective, but with the additional uses for the scout, the economics look better than ever.
He estimates that a scout will sell for between $75,000 and $80,000. Whereas the larger growers might own one or even several scouts, he thinks packers might purchase scouts for their growers to use. Agricultural service companies might also buy scouts to rent out to growers.
Operating expenses will be low, particularly if the scout is towed autonomously and doesn’t need a driver, Morikawa said. "It’s mostly just the cost of the equipment. Field costs will be negligible."
Work is under way on the robotic harvester, also. Students at the Franklin W. Olin College of Engineering in Massachusetts are developing picking hands for both the apple harvester and a citrus harvester, as a senior project. Morikawa said this is one of the most challenging aspects of the apple harvester, as the hands must not bruise the fruit while picking it. Technology for guiding the picking arms will be similar to that of the vineyard pruner that Vision Robotics is developing. Robotics might sound futuristic, but Morikawa said the scout project is doable.