There is probably a sizable market for building out software and technology for 'upgrading' older farm equipment to modern principals, such as self driving, variable rate fertilization, variable rate planting, etc. When I was more involved in ag research I met several farmers who DIY built themselves variable rate fertilization drills. While its not ubiquitous, the farming community has a very long culture of DIY out of pure necessity. I think this is the source of the pushback against companies like JD.
>> the farming community has a very long culture of DIY out of pure necessity
Can confirm. And it's largely because a farmer's job is the management of dozens, hundreds, thousands, or tens of thousands of living beings - whether those are animals or plants.
Thusly the level of unpredictability exponentially increases in this field versus a number of other fields - and yet this field is hypercritical to the sustainment of humanity itself.
With weather, disease, and all other mitigating factors taken into mind, and despite a farmer's best efforts, there may be entire large sections of crops that completely fail - or animals that get diseased and affect the rest of the herd. This can lead to significant losses that I've unfortunately experienced firsthand.
If it's Harvest season - that season only lasts a very certain, specified period of time, and if equipment fails, farmers will do whatever they have to do to hack around to make it work.
I've seen an incredible amount of DIY innovation firsthand in these kind of crises. Farmers are some of the most hardworking, dedicated, and innovative people I've ever had the pleasure to meet.
And let me tell you - when farmers are dealing with crises like this - the idea of a piece of software getting in their way, when a single day or two can make all the difference in the world - it's not something they want to be involved with, unless they know they can fix it themselves.
I've heard of farmers going so far as to make their own GPS guidance systems, enhanced before they were widely available and reasonably priced. I'm not sure how accurate and repeatable they were.
Going to any farm that's been around for 20+ years, you'd see so many custom solutions that are really functional.
Farmer one-off solutions are kind of like homebrew Excel applications - they're infinitely customizable, the end user knows exactly how it should work and can reasonably make it happen, and it's a small fraction of the cost of a commercial/IT SOLUTION. It may not be as reliable, but it's good enough.
Trimble GPS receivers go for around $10k (USD) [0], so I'm not surprised farmers try to build their own devices. Here's one who built an RTK GPS system out of a Raspberry Pi (and various other parts) for £250: https://www.youtube.com/watch?v=Kwxpo04AC5Q
Not by itself, in most situations. Maybe if you're automating a combine, where you're doing repeating patterns in a big flat field with no obstacles and where 1 meter deviations from the desired trajectory isn't an issue (not an uncommon situation, but for a lot of use cases you really do care about that 1 meter deviation, e.g. for plowing or cultivating).
Otherwise you have to augment, eg with relative GPS or machine vision.
I belive they were enhancing it with terrestrial base stations (Loran-C was mentioned in the discussion, which was 10+ years ago). John Deere has proprietary secret sauce and if I recall correctly, they can get sub 1" accuracy. That precision farming has enabled things like different amounts of fertilizer for different parts of the field, a substantial material savings and improvement for the environment.
u-blox's ZED-F9P chip is capable of 10 mm (0.39 inch) accuracy [0]. AFAIK, high accuracy requires being relatively close to a terrestrial correction station. Some states, such as Iowa, run a network of these stations and provide free access to the public [1].
Everything that you mention except the self-steering is a function of the attached implement, not the tractor. That makes incremental upgrades of that sort much easier.
Well, no. Precision agriculture is an integration of many components of the farming operation, all of them very much data driven.
Lets just take the example of variable rate fertilization, because its the one I'm most familiar with. In the case of winter wheat (hard red and white), there are 3 primary components determining both yield (kg/m^2) and protein (g/kg). They are the available water, the available nitrogen and the 'site index' which is generally related to soil depth, but slope and aspect also play into this.
So, if what you want to do is land on a very specific amount of protein, say 7.5% across all of your fields in a dryland cropping situation. You need to know how much much water is available at all points in your field; how much nitrogen is available at all points in your field; and what scalar to carry around for a given XY position in your field. Maybe think of the scalar as an 'X-factor' to multiply the result of Water*Nitrogen by to adjust for local conditions.
Wheat, grown in high water low nitrogen gets very tall and has low protein and higher yield (to an extent). Wheat grown in high nitrogen stays short, but has higher protein and lower yield (to an extent). So it becomes a trade off between protein % and yield, and the theoretical maximum is described by the site index.
Wrapping it all together. First, find a cooperative old kook of a farmer. Second, convince him to let you drill a hole in his up auger to mount a hyper spectral sensor. Third, duct tape/ bungee mount a LiDAR puck to the front of the combine. 4th, solder some leads into the on harvest yield monitor. 5th, mount a high quality GPS unit to the top of the harvester if you don't have one already (or can't access the NMEA stream). Add car batteries, some old notebook computers to get the data, and a few days to harvest. Boom. Now your combine is a fully integrated data collection platform. You now have a measurement of grain height at all locations in your field. You also have a measurement of protein content at all locations in your field (with a little PCA wizardry and some protein sampling). Combine the two and you now have a 'site-index' that is nitrogen/ available water independent that you can use in your variable rate drill to adjust the fertilization rate to match the site index (since you can control the fertilization rate and not the available water). This will give you fairly precise control and allow you to more or less homogenize the overall protein content of the field. Alternatively, you could also optimize for yield (allowing protein to fluctuate).
