Cover image: The OMC Shockwave X is an autonomous shaker that doubles
the shaking speed without an operator.
With a goal to automate the orchard, this agtech company is tackling the tough stuff first.
As an agriculture engineer who designed tree shakers and then wrote the software to control them, Tyler Niday, CEO of Bonsai Robotics, had always wanted to found a startup. He also had an affinity for the tree crop industry.
He started with yield monitoring, but realized there was a bigger problem inside the industry where precision agriculture hadn’t taken off quite like it had in other crops. The machines got more mechanized and more complex, but with increased utilization.
“But there isn’t the same precision ag revolution that happened in corn, cotton, and soy,” Niday said.
Niday mentions this in regards to tech such as GPS autosteer, variable rate spray maps, and yield monitoring and forecasting.
“We set out to build the tech stack to accommodate the challenging conditions for the orchard space,” he said, such as for GPS-denied areas, dusty environments, and other conditions that are difficult for autonomous vehicle perception.
Bonsai was built around this opportunity in harvest in particular, to partner with the people and companies who are good at building machinery, and with Bonsai building the software.
The Machinery
Instead of off-season or more tractor-based autonomy, Bonsai started with harvest – a difficult and riskier one to tackle because of the dust, debris, and the increase in actions, such as interacting with the trees specifically.
This solution is a camera-based system instead of LIDAR, a difference from other autonomy.
“I’m really a fan of computer vision, but if you think about autonomy being commoditized, a camera is the most affordable sensor. And cameras see in color, so there’s a lot of contextual information to base things off of to make the right decision,” said Niday.
When that 3D tree map is created, there is more information available outside of driving and navigating. New features will be made available to offer better operation and insights into per tree health, such as selective spraying or per tree yield monitoring.
When there’s 30 megapixels of cameras and taking in info from 360 degrees, there’s a lot of insights growers can get with applications in the field.
The opportunity is to optimize outputs with the same sun and soil, but with a paradigm shift of managing fields to manage the trees. And Bonsai doesn’t have to add much additional hardware in order to make the machines autonomous.
“They have great muscles and ligaments, and we attach the brains and eyes,” said Niday.
Bonsai is currently partnered with Flory Industries, Orchard Machinery Corporation (OMC), and a parts manufacturer.
The first product was in the OMC Shockwave X tree shaker (cover image). The focus for the initial offerings were products that weren’t simply labor replacements, but also increased speed and utilization.
For example, going from shaking four trees per minute to eight trees per minute is a huge increase in ROI to the grower. This particular shaker is a whole new design, and it doesn’t stop at each tree, but continues to move from tree to tree in one continuous motion.
In Flory equipment, Bonsai has VisionSteer in the single-pass V62 sweeper, which is a 3-in-1 machine that’s large and can be difficult to maneuver for newer operators. Vision-assisted steering helps during harvest. Outside of harvest, the sweeper can be removed, and the machine can be used for spraying and mowing applications.
“We really targeted these applications — multipurpose machines and increased speed, in addition to an operator assistant or replacement,” said Niday.
It increases speed and enables any operator to run it.
2D to make 3D
In traditional autonomy, a variety of sensors are placed on a machine and fused together to start with a 3D view of the world. But, that 3D view can be jeopardized, such as by debris or dust clouds in front of the sensors, making that view of the dust cloud instead of the trees around the machine.
A human wouldn’t drive based on this dust cloud, but would drive through it. And this is, in its most basic terms, how Bonsai tech operates — like a human. The cameras are the eyes, and from those cameras run AI models that produce tons of output in 2D. The system knows if it’s a young tree, a dead tree, the size, the angles, and if the area in front is drivable.
The 3D world around the machine is created from those 2D images, allowing the machine to position itself accurately within that 3D space. So in the dust scenario, the machine can detect past the dust.
“Normally, you would use GPS to get your location, but we do it all from vision,” Niday said. “2D image in, 3D world out.”
Bonsai’s goal is to give every tree the “master Bonsai treatment.” Not to make a small tree, of course, but to put so much effort into each and every tree that it can multiply yields.
Grower and OEM influence
“The only way you can have a product be successful is if there’s an actual return on investment for a grower,” said Niday.
Niday says growers should see somewhere around 45-50% savings over the four-year period a machine would be amortized. There is an activation and subscription option, but for those who would rather pay up front and receive the service at a discount, that’s also available. The service provides free over-the-air updates any time there’s an improvement or new feature. There is also full tracing and telemetry, so growers know where a machine was, what it did, and additional insights.
Bonsai has relied heavily on feedback from growers and the expertise of the original equipment manufacturers (OEM) with which it partners. These same OEMs have been making significant strides in their own innovations and advancements.
“We help each other, and when you pair it together, it’s awesome,” Niday said.
Photos courtesy of Bonsai Robotics
Other Articles
Positive results with in-store almond campaign
