32
votes
Robot that uses AI to pull weeds may reduce poisonous herbicide use by 70%
Link information
This data is scraped automatically and may be incorrect.
- Title
- Mother plucker: Steel fingers guided by AI pluck weeds rapidly and autonomously
- Authors
- Benj Edwards
- Published
- Nov 28 2023
- Word count
- 497 words
I just love the idea of thinking what weed evolution will look like when the successful outcome involves tricking the AI. Weeds that look increasingly like the plants they're surrounded by? Small weeds that hide in the shadow of the stalks of corn? It's kinda neat
Food for thought.
That's a new page to add to my favorites
Oh yeah, I knew this was already a thing but it's fun to ponder what would change. There are always unintended consequences
The robots just need gustatory sensors so the selective pressure causes weeds to taste like corn? Like chicken (though I'm team 'default tetrapod flavoring').
That would be a neat outcome. If they taste like corn, we can keep them!
I realize this is new technology and thus won't hold up to current methods of weed control
But 25 acres a day is miniscule compared to current methods. I can spray ten times that in about four hours using a conventional high speed sprayer.
If you could keep them running 24/7 would it matter? Especially if they could be charged using solar or something.
Well the main issue is that you'd have to have a small army of these things to be effective. Weeds have a tendency to grow, so the best time to eliminate them is early in the season and well before they have an opportunity to reproduce. For anyone doing much more than a small (<100 acre) hobby farm, 25 acres a day just won't cut it.
There's really only a window of about 2-4 weeks during the season where you can eliminate weeds before they start to affect the growth of your crops and also before you're causing damage to the crops you want to grow (by driving over them)
I'm possibly misunderstanding your argument, but if you have 2 weeks to eliminate weeds, and this device can manage 25 acres per day, then you can accomplish 350 acres in that time period. As far as my research tells me that is about the average size of a typical organic farm in the U.S.. If you have 4 weeks and only need a single pass you could manage 700 acres with a single device and a single battery.
That said, I have a feeling this is going to be desired by small scale organic farmers for the time being, if at all.
I can't speak for organic farmers (because I'm not) but I can say that for standard farming operations in my location, 2500 acres is an average sized operation. We have more than one location larger than 300 acres.
I can't speculate on whether one pass with this would eliminate a sufficient amount of weeds, but I can say that our weed prevention program involves a burn-down spray before the seeds get in the ground followed by a secondary herbicide spray somewhere between the crops emerging and the crops being able to shade out competing plants (weeds) on their own
I suppose you could have a small army of these drones for the season. If that's too costly for an individual farm, someone centrally located to multiple farms could have the ground crew and rent them to each farm over the course of the two weeks.
In other words, it'd need to be cheaper than day labor to be economically sound.
The idea is to reduce use of harmful chemicals, there are going to be tradeoffs. That being said, 25 acres a day is pretty good considering its autonomous. Sure you can do it faster with a sprayer, but now you don't have to worry about it.
For sure. It is just an early model; it's not gonna be competitive yet.
There are somewhat similar improvements being made to limit how much a sprayer has to spray to be effective. I know a few companies are developing technology that will turn sections of the spray boom on or off depending on whether cameras sense a weed while you're driving through the field.
Less area to spray = less chemical to spray = less money spent on chemicals for that year
Catch is it's really prohibitively expensive at present. (Looking at you, John Deere....)
It also requires computer support and is likely fragile over time. How practical and cost effective it will be to implement ai, especially in low margin business like farming is a very good question. Servers are not free and not without environmental cost.
This robot won't be using any servers. It will have on-board graphics cards that can do the computer vision on the fly.
I wonder how much knowledge has transferred from Battle Bots / robot wars. When can we put lasers on our robot mowers to weed so we don't need to rely on fertilizer?