Tractors do a dance. They go up a row, offset their track by a dozen or so feet, go down a row. One tractor might be for tilling, another for harvesting, another for trucking the harvest to the bunker to be stored. They're all following an orchestrated set of rules. But with humans behind the wheels and entropy at play, it's not as simple as till a row, sow a row. Machinery breaks down, and parts of crops are lost to rain, sun, and general inefficiency.
If you ask a farmer tomorrow how his harvest efficiency is doing, he may not answer that question. Until you know you need it, you don't much miss it. "He might look at you funny," says Brian Luck, professor of Machinery Systems and Precision Agriculture Biological Systems Engineering at the University of Wisconsin-Madison.
Something most can agree on, however, is the need to get as much of a crop off of the field as possible, at the highest quality possible.
Luck and his students in the Biological Systems Engineering program are working on ways to improve how the whole farm's run. In recent years, they've developed Unmanned Aerial Vehicles that use thermal sensors and infrared wavelengths to read the moisture of a corn stalk or how heavily it respires. But drones used for agriculture are still quite futuristic for the typical farm. For optimizing harvests, you'll need to start on the ground.
One of Luck's graduate students, Joshua Harmon, is doing his thesis work in forage harvest logistics: How the stuff we feed to dairy cows and cattle is harvested, and how GPS and CAN (Controller Area Network) systems can streamline the process.
How it works
Harmon describes the tractors' dance more like they're running lines of code. Data from CAN—the same kind of system that turns on the check engine light in your car—shows what the transport vehicle is doing, while GPS data shows where it was doing it. Newer vehicles have these systems built in, but for the older tractors, he retrofitted them to include GPS and CAN data collection.
"For any given point of time, a harvester was constrained to the given work states: harvest, travel, delay, idle, and downtime," Harmon explained via email. The states are defined as:
Harvest: Exists when the harvester is actively loading a transport
Travel: Movement to fields, as well as headlands turns
Delay: Time spent delayed due to detecting metal (forage harvesters have metal detectors, and the rolls drawing feed into the machine shut off to prevent knife damage if something like a fencing nail were to get picked up)
Idle: Time spent waiting on transports
Downtime: Time spent broken down or other reasons for stopping the operation
From there, the transports each exist in their own states: loading a harvest actively, travelling to the storage site or field, unloading at the storage site, or idle.
By watching how these states cycle as the machines go about their field business, they can start seeing patterns and gaps in efficiency. "Based on all of that, he could essentially define cycle times for how long it took a certain type of truck to get loaded, how long it took to get back to the bunker and unload the vehicle… It defines the whole cycle process," Luck says.
From there, it becomes a word problem: If I have a forage harvester with "X" harvesting capacity and my field is "X" distance away from the storage site, and I know all of the routes I have to take to get to the field, what is the optimal number of trucks that I need to maintain harvest efficiency and never have that forage harvester stop and wait on a truck to get back?
The solution depends on never having a harvester waiting on a tractor. As a farmer, you want a just-right number of tractors running at once. Too many and you're wasting manpower and machine-power (which equals money), too few and you're wasting your crop yield, a.k.a more money. Precious time is gained when drivers know exactly which field entrance to use, where in the field the harvester is, and most importantly, when it needs the next empty truck to fill.
Big data, big business
You might get a funny look from a farmer about efficiency, but it's a problem agricultural scientists have been working on for decades. Some suggest new strategies for timing, others use tech to track satellite imagery and weather data for a better yield. It's studied for good reason: Forage harvest, and silage, is a big part of the industry.
Harmon's work is focused on silage, a livestock food that's used when forage production gets low. Wisconsin produced nearly 19 million tons of corn for silage in 2015, and in 2012, the U.S. corn industry was valued at $67.3 billion, including both corn for grain and corn for silage. That's 17 percent of all agricultural sales in the country, and Wisconsin is king of the silage states. It's big business, and big data could help it boom even more.
But there's not a set standard for how efficient a farm needs to be to succeed. "The ultimate goal is to provide the producer with the tools necessary to enable the most efficient harvest possible," says Luck. There's a balance between paying an extra tractor driver at the ready, and improving harvest by a small margin.
"Making silage production more efficient allows producers to get more work done in a shorter period of time," Harmon says. "As a result, silage quality improves, because there is a better chance of harvesting at the right maturity and moisture, as well as less chance of being affected by adverse weather conditions."
Once you've broken the goings-on of a harvest down to states and statistics, it makes one wonder: Are people the biggest problem with farming efficiency?
"People are talking about self-driving semis, and self-driving cars… I see a lot of risk involved with that, when you're talking about somebody riding in the vehicle and not driving it," Luck says. "It's all well and good—I'm gonna be one of the early adopters of that, I think—but I see a greater opportunity to at least perfect the technology [in agriculture]."
He's not alone in the idea. The Case IH autonomous tractor drew skeptics and admirers at the September Farm Progress Show in Iowa. Like Harmon's systems sensing machinery's position in the field, a farmer in Maryland is using GPS to guide his own self-driving John Deeres. Slower speeds, simpler parameters and less human interference make the farm field a playground for robot tractors. It's the difference between learning to drive in an empty parking lot and a four-lane highway. "Probably in 10 or 20 years, I think we will have some sort of autonomous field operation machine that have been implemented and being used," Luck predicts.