Manitoba wheat farmer Matt Reiser has turned a tractor into a fully autonomous drone using little more than a Pixhawk and his own ingenuity. His automated tractor has saved him more than $5,000 in labor costs this year, but it hasn’t eliminated any jobs on the farm — instead, it’s made the workers he already employs more productive. In fact, when I first spoke with Matt about his project in August, his harvest for the year was already 75% done, which is unusually early. He attributes some of that to his new robot.
“I actually can’t believe it worked,” says Matt. “I’m as surprised as anybody.”
Behind this humility is a story of skill, risk, tenacity and technology, from an unexpected place — the expanse of the winter wheat farms of mid-western Canada.
Though Matt didn’t start his robot tractor project until just this past winter, as far back as three years ago he’d had the idea that an autonomous vehicle could handle the rote work of operating a manned grain cart. It’s an easy job. During harvest, you need someone to drive the cart up to the combine about every fifteen minutes so the combine can unload grain. The whole operation takes about five minutes. It’s such a simple job that Matt often has someone on the grain cart who isn’t even familiar with tractors; it’s only a job for a couple weeks a year.
Like many farmers, Matt had already been using tractors that have an autosteer function. Autosteer assists the driver by keeping the vehicles moving forward and back on a straight line. Matt didn’t think it’d be too big a leap from there to being able to send the cart automatically to specified points in his field, but he didn’t know where to start.
Then last year Matt came into a little Christmas money. He wanted to put it towards an RC plane, a hobby he’d had as a kid. Matt started googling and came across something new in the RC world: autopilots. When he saw the Pixhawk he realized this was the solution he’d been looking for: for less than $300 he could put a Pixhawk in his grain cart and essentially turn his tractor into a robot.
This all proved a little harder than Matt first planned on.
“I knew nothing!” Matt says. “The Pixhawk had the communication all set, but I didn’t know anything about the architecture of the developer streams.” But he soon discovered DIYDrones.com and the ArduPilot community, whose ArduRover firmware was perfectly suited to his project.
“The community was really great, especially Philip Rowse and Grant Morphett,” Matt says. “My first post on DIYD was like, hey guys, I’d like to set this thing up. But I didn’t get much of a response because people didn’t know if it was even going to work. Once they saw it was going to work they were more than happy to help.”
The community could also act quickly to fix bugs. “There was this software problem, where if the radio lost connection and the vehicle was driving around it wouldn’t stop,” Matt explains. “Turns out it was a bug in the Rover code, but the community was on it, and less than 24 hours later Rover 2.5 had solved the issue. They also helped me out with these basic problems that RC experts would never have had in the first place. They were so patient.”
So was Matt: “I realized pretty quickly that this wasn’t going to work exactly, so I had to learn some programming skills.” He enrolled in an online coding course and learned how to write Python.
In applying the technology to the tractor, Matt first had to program the steering. “I thought that was going to be hard, but it turned out to be really easy.” He ran the signals from the Pixhawk into a board with a motor controller, which operates a hydraulic valve that’s already on the tractor. It sends 12 V to one side to turn the tractor one way, and sends it to the other side to turn the other way.
So the tractor steered, but with no feedback it only drove right/left, right/left, in a straight but kind of wavering line. Matt installed an angle sensor, so the Pixhawk could detect where the wheels were pointing, and after that the tractor maneuvered pretty well.
For speed control he made a couple of brackets for some linear actuators. “It doesn’t even need a clutch; it’s all controlled by the onboard computer. Throw it into forward and the tractor goes forward; put it in neutral and it’s like stop. The Pixhawk operates all of the physical controls.”
With the mechanics down, Matt needed an interface so an operator could control the tractor remotely. He applied his Python programming knowledge to writing an ArduRover interface using the DroneAPI from the 3DR site. “The interface looks like a kid with a crayon drew it,” Matt says. “It’s just a bunch of big colorful buttons. The guy who operates the grain cart is 60 years old. He had a hard time figuring out GPS, but has no problem using this.”
In the field
Of his first year of autonomy, Matt says, “It’s been huge! W’ve been getting way more done.”
When the combine operator sees that the combine hopper is full, he pushes a button on a computer, and the button says “start unloading.” The grain cart automatically comes to wherever the combine is, lines itself up, turns and drives right under the unloading bay, so the grain drops into the cart. The grain cart has four times the size of the combine’s capacity, so the combine can empty itself four times into the cart while it’s still moving and harvesting. The operator then just pushes the “empty” button and the cart pulls away, turns around and waits for Matt’s emptying cart to come over. “I can be running back and forth to grain storage and always keep the grain cart empty, so we never have to stop combining.”
“Sometimes it’s a little scary having a person drive the cart. Scarier than having the Pixhawk do it, for sure.” Matt’s devised a number of failsafes to stop the tractor — for instance, he can just pull the USB out of the computer and the grain cart stops. “Now the cart driver can do other jobs around the farm. So we’re doing more.
“The other thing we’ve saved, is we’re spilling less grain,” Matt adds. “When the Pixhawk drives, it’s a nice straight line, so nothing spills. No human error. It doesn’t wobble.”
Matt’s innovation appeals to neighboring farmers, and he sees increased autonomy as heralding an enormous change for the agricultural industry. In fact, he figures he’ll soon be able to sell his system to other farmers around him. And who knows, since he’s so far out in front of adoption, he might even be able to quit his day job, or at least hand it over.
I’ve been showing to everyone, neighbors and friends. Definitely recommend — . Plug into data port of tractor. Lots of interest. Big farm show in fall — Ag Days — speaking there this fall about building and open source.
His work has generated a lot of interest. His YouTube video of the tractor got over 100,000 hits, and he thinks he can sell some of his kits over the winter. He admits that some of the technical aspects are probably more than most farmers want to deal with, but he doesn’t anticipate that it will be too hard to integrate with the electronics, especially with tractors that already have autosteer. He’s speaking at a large farm show this fall — Ag Days — about building and open source technology. “It’s going to change farming,” he says. “This autonomous tractor stuff was coming, but what I’m doing will save money.
“I really want to express my gratitude to 3DR and everyone in the community who’s worked on these projects,” Matt says. “The ArduRover code, MAVproxy and Mavlink — I never would have been able to do any of this if it weren’t for the support of that community. They made it happen for me.”