This seems like kind of a fluff piece, I don’t see any details about the actual technology. The whole article is about how proud the company is to be partnering with midwestern farmers. Can anyone provide a link or more information about how they are able to distinguish weed from crops, and by what mechanical process the weeds are removed?
But the technology is: A library of weed images is used to check if the thing visual to the camera is a weed. If it is, then pluck it or spray chemicals (here: pluck).If not, move on.
Same I can’t find any details on the mechanical mechanism they remove the weeds with. That seems like as much if not bigger challenge than identifying the weeds.
I don't think the mechanical mechanism is that hard.
Once they detect a weed ahead, they could just lower a small plough-like thing in worst case scenario. Or a pinch could go down, grab the weed and go up.
But yes, they seem be secretive about the working.
Deere bought Blue River, which does this. Deere packages it as a wide implement attached to a tractor. Cameras look down, recognize weeds, and zap them with targeted weed killer. About four companies are already in this space.
It's sold as "weeding as a service". If you need large-scale weeding, Pacific Ag Rentals in Salinas, CA, will rent you a Robovator. That's made in Denmark, and Pacific Ag Rentals beefed it up for serious use. It kills weeds by chopping them out of the ground with computer-controlled knives. No chemicals, so, "organic".
Making these things rugged enough to do the job is a problem. The R&D projects tend to have flimsy mechanical engineering. You want to be able to pressure-wash the thing. Dust-tight and water-tight are solved problems, but most machine learning people have never dealt with that.
I've been going to my local farmers' market for a while. It's interesting to see what people use "organic" as a proxy for. For some, they want to support small family farms. Others will settle for "no spray" produce that's not certified organic. Some people are scared to death of GMOs, and that's one way to avoid them. Organic eggs are also free-range (but it' still not as nice as you'd picture), and there's minimal use of antibiotics during production.
This tech seems to have the potential to decentralize farming somewhat. The same machine with a few attachments and software options could also cultivate, plant, irrigate, debug (organically with frikin' laser beams), and harvest. Since software is the most expensive part of the problem it could be turned into a small retail device in not too many years. Buy a gardenbot like you now buy a lawnmower. Add seeds, compost, power and water, let it loose in the backyard and just wait for it to stack vegetables in a bin.
Machinery is pretty expensive in its own right. Also tractors already have the ability to pull multiple different implements behind them. The only thing novel about this is that it's somewhat autonomous. Even that isn't terribly novel since most large scale farming uses GPS guided steering at this point. Ag is very very high-tech already.
The automated equipment is still so expensive hardly any of it is owned or used by family farms. The thing was that 10+ years ago the automated equipment was owned by very large ventures (large cooperatives and bigger) that shared or sold service only, traveling the region with the equipment to sell that service. Pretty sure that case has hardly changed. Travel the country and you see that most farms don't have the latest equipment rolling or sitting on their land.
what leads you to think that SW is the most expensive part?
IMHO it depends on scaling- as SW scales extremely well it might be, that SW is by far the cheapest component in this mechatronic system.
especially if you factor in maintenance and operating costs.
There are a lot of companies working on something similar. But I hate to disappoint everyone thinking that this will mean the end of herbicides, it will not.
But there are a lot of crops, vegetables mostly, that are hand weeded. It's going to replace a lot of migrant labor. The main reason it won't replace herbicides will be cost. I'm not even certain that it would make economic sense right now for organically grown corn and soybeans.
There also may be problems with it in the Midwest. In California they turn on the water when they need it. What happens if it rains for three weeks in Illinois and the machine can't enter the field? Will it be able to remove fifteen inch tall weeds as easily as those a quarter inch tall? How fast will it operate under those conditions? If I have 2500 acres in that situation the weeds might be five feet tall by the time the machine gets to the last fields.
But if someone makes a successful business of commercial weed removal then like all tech the cost will come down over time, perhaps opening up the much wider market.
Absolutely we should expect them to solve easy/valuable problems first. Maybe that's fancy organic vegetables in a greenhouse, which certainly sounds much easier than 5-foot weeds growing in a thousand acres of deep mud!
But we farmed for a long time before inventing weedkiller. First by recognising weeds & pulling them out. Later by plowing, which is really just substituting mechanical work (from oxen) for human labor (and smarts). I don't see a fundamental reason that robot-weeding couldn't ultimately replace most chemicals & most plowing, but obviously not next year.
I'm still waiting for the version of this that removes invasive crops from non farm land. So DOT could unleash it on any road side, removing kudzu etc and allowing natural plants to take back over. Or under power lines, only letting trees with a max 10 - 15 ft height survive so further maintenance is reduced in the long term.
There are some incredible maker-farmers involved in ROS. Not downplaying the people and tech in this article but farmboys with no formal education are building computers and microcontrollers run by ROS that strap onto 30 year old farm equipment and make it autonomous.
How is this machine powered? I did not see a bank of solar panels. So are they plugging it in at night? Fuel costs (electric or chemical) would, I think, make this a hard sell against large spray boom operations.
Is anyone else concerned that food production is being consolidated into the hands of a few massive corporations? Somehow this is more troubling to me than the energy, automobile, or computer industries
23 comments
[ 4.1 ms ] story [ 62.3 ms ] threadBut the technology is: A library of weed images is used to check if the thing visual to the camera is a weed. If it is, then pluck it or spray chemicals (here: pluck).If not, move on.
Once they detect a weed ahead, they could just lower a small plough-like thing in worst case scenario. Or a pinch could go down, grab the weed and go up.
But yes, they seem be secretive about the working.
It's sold as "weeding as a service". If you need large-scale weeding, Pacific Ag Rentals in Salinas, CA, will rent you a Robovator. That's made in Denmark, and Pacific Ag Rentals beefed it up for serious use. It kills weeds by chopping them out of the ground with computer-controlled knives. No chemicals, so, "organic".
Making these things rugged enough to do the job is a problem. The R&D projects tend to have flimsy mechanical engineering. You want to be able to pressure-wash the thing. Dust-tight and water-tight are solved problems, but most machine learning people have never dealt with that.
I've been going to my local farmers' market for a while. It's interesting to see what people use "organic" as a proxy for. For some, they want to support small family farms. Others will settle for "no spray" produce that's not certified organic. Some people are scared to death of GMOs, and that's one way to avoid them. Organic eggs are also free-range (but it' still not as nice as you'd picture), and there's minimal use of antibiotics during production.
But there are a lot of crops, vegetables mostly, that are hand weeded. It's going to replace a lot of migrant labor. The main reason it won't replace herbicides will be cost. I'm not even certain that it would make economic sense right now for organically grown corn and soybeans.
There also may be problems with it in the Midwest. In California they turn on the water when they need it. What happens if it rains for three weeks in Illinois and the machine can't enter the field? Will it be able to remove fifteen inch tall weeds as easily as those a quarter inch tall? How fast will it operate under those conditions? If I have 2500 acres in that situation the weeds might be five feet tall by the time the machine gets to the last fields.
But if someone makes a successful business of commercial weed removal then like all tech the cost will come down over time, perhaps opening up the much wider market.
But we farmed for a long time before inventing weedkiller. First by recognising weeds & pulling them out. Later by plowing, which is really just substituting mechanical work (from oxen) for human labor (and smarts). I don't see a fundamental reason that robot-weeding couldn't ultimately replace most chemicals & most plowing, but obviously not next year.