Inspiration For Weed Killing Robot Tool

I wanted to suggest this as I believe the best kind of weeding tool should be mechanic and not chemical. Makes sense economically and environmentally (plus a health bonus too I guess).
In this link they describe a system that uses a sensor to identify weeds that then get bashed to pieces! Seems pretty neat so maybe something similar could be useful to you.

Thanks for sharing! I agree that the best way to remove a weed is via mechanical disruption/destruction. As long as weeds are detected when they are sufficiently small, then it should be fairly easy to push them into the ground, cut them up, or use an auger type drill to destroy them. Using chemicals is unnecessary if you can detect early and often!

1 Like

I’m a farmer. Weeding is super easy. You just have it do a 48 hour auto-weed. All you need is a bent wire that looks like a stirrup. Different plant types have different root types so proximity to seeded coordinate can be based on plant type.

The problem on our end with weeding is the timing and labor. If ANYTHING is more important in this project to understand than this in terms of what this device/system could achieve it is that. I understand the there are problems to overcome in the design but in the end, weeding and water use are going to be the two most fundamental reasons why we farmers need this device.

Hello Everybody , My name is Jabbar and first I would like to thank you for opening such a door so that people can share their ideas for a better food future.

I am a professional in the agriculture industry and enthusiast of IoT.

I found that link useful and I thought it might be useful:

Unfortunately , it is in French but I think everybody could understand.It is about a robot that is created for spraying herbicides with precision on agriculture field.

Well, real weeding, like a human can do it, is probably far away(because of complicated imageprocessing and fine scale mechanic), but basic weeding in an early stage shouldn’t be too hard, with the current set-up.
All you need is a camera on top.

Because you know, where your plants are … so everywhere else where green on the field shows up, you push in, unless its in the area, where your seeds are.
And with time you have to increase the nopush-radius (or do so automatically, with a little bit more complicated algorithm, who searchs for green connected to your seed area).

So this could get you over the critical first time, when weeds will overgrow your plants.

I realy like to try that approach out for myself, but unfortunately I lack the money and time right now … but maybe someone else can try that out.

Yep, that’s exactly how we’re approaching weeding. If you can get them early and often, then they will not overgrow the plants you want. And if you do this season after season, then weeds will become less of an issue over time. We’ll be demonstrating weed detection and removal on July 1st!

So you stole my idea!
… well, more seriously, I would have been worried, if you didn’t figured that out by yourself :wink:
I just have not seen it described somewhere, so I thought my input wouldn’t hurt.
Good luck and nice to see some progress

Haha thanks!

1 Like

I ma curious as to how well their current 4-prong weeder is working. I use a stirrup hoe (see: like what you are talking about, and find it insanely fast and easy.

The issue that I see with trying to use one on FarmBot is the lateral stresses on the gantry. I am wondering if the current weeded could be modified with a stirrup hoe add-on. (you may need to make 3 weeders, which can be mounted so that stirrup hoe part is 120 degrees apart. The weeding function would then be:

  1. weed is detected. Farmbot identifies a “safe” place nearby (2" away) to :land" the weeder. If the orientation requires a different weeder orientation, it swaps to the right one.

  2. By descending on the Z access, Farmbot “lands” the weeded head, which goes (perhaps 1") into the soil, before further penetration is stopped by a plate on the prongs. This secures the bottom end of the Z arm, and allows for some minor torque to be applied.

  3. As the Z arm descends a bit further, the downward moment is converted by a cam to a lateral movement, which would pull a small stirrup hoe through the stem of the weed. Since this cuts off the entire plant above ground level, it is VERY effective. The stirrup hoe only needs to move 1/2" or so.

  4. As FarmBot raises the Z arm, the stirrup hoe is reset (by a spring) for the next weeding action.

This has the added advantage of allowing weeding to happen closer to desired plants than the current weeder.

That sounds like an awesome idea! The current weeding design works, but it is very crude and “brute force” in its nature. Have something a little more elegant and tactical may work better. Do you have any experience with CAD? Maybe you could draw up an initial design!

1 Like

I don’t personally have any experience with SolidWorks. However, the company I work at DOES have a SolidWorks designer (currently on holidays). If you can provide an email address that I can communicate to privately (outside of the forum) I will give you my email address. If you could send me the file for the current 4-prong one, I could see if he would be prepared to modify it.

I spoke with one of our designers today, and she is prepared to do it (I’m going to pay her a bit.)

Do you think that we could use this ( as a cheap, standard, replaceable blade? The blade WOULD be out in the elements all the time, so would need to be replaced every month or so, but $1.49 for a pack of 10 is pretty cheap.

Looking at the knife blade gave me a different idea.

Suspend a holder for a scraper blade 2-3" under the tool head. (This can just be a simple support slab on each side.) The blade is pointed down by maybe 15 degrees, so that the sharp edge is the lowest point. Put a small micro-switch on the bottom. Potentially add a safety / environmental protection shield to the front.


FarmBot detects weed.

FarmBot activates power to (small motor? electromagnet? other?) and shield swings up and clear.

FarmBot lowers head 1/2" away from weed, with cutting edge facing weed.

Farmbot raises head 1/32", so that there is no ground contact. (reported through data port.)

Farmbot advances to touch weed with blade, then lowers shield to pinch and cut off weed, then carries weed to disposal / collection location


I can draw this up as a 3 view 2D drawing if you want. Certainly open to suggestions.

The weeder, just stabbing the weed can break the weed up and each part can then propagate, making you weed problem worse. You need pincers or a rotary weeder to pull the weed in whole, then dump in a bin.

GordonHolmes ever head of the process called ring banking with sucker bashing. Its a method for exterminating trees without chemicals.

Stabbing and breaking the weed apart yes risks that each part can propagate. The key thing to non chemical bashing is that you have fairly much have to do a pass every few days to get on top of what ever propagates from being bashed.

Plant need light to make chemicals to grow leaves. If you keep on bashing it regularly enough the weed runs out of fuel might propagate a few times before it absolutely dies except for some insanely determinate plants. This is only one method of exterminating weeds without chemicals and without pulling them out of course having more than 1 method is a good idea. Even pulling them out can snap off roots and you have the root propagate issues. One that comes to mind that resists being pulled out and bashed is nut grass.

To kill something like nut grass environmentally is water it with boiling hot water(and I do mean boiling as close to 100 C as you can get) after its watered you can dig the top in. In fact for weeds that leaves spread water ways from them bash top in then water with boiling hot water. Reason for the boiling hot water not just a flame touch or steam is it goes quite deep into the soil profile does not just kill the bits on the surface also killed everything that it cooks under the surface.

Boiling hot water weed control is not classed as that power effective. Weed killing effectiveness boiling water beats most chemical weed control.

Combination to bash weed then coat in boiling hot water will against most weeds stop any part of the weed being able to propagate. Any of the weeds not killed by this process will in most cases live through pincers or rotary weed extraction. The three biggest issue with the boiling hot water

  1. It not absolutely selective there might be cases where you treat a weed with hot water and the plant near it dies.
  2. hot water risks human using it getting burnt. Integrated into machine would allow it to be used safely.
  3. power usage.
    So just like chemicals you would want to use boiling water methods with restraint the upside is that boiling water is one of the most effective killers out there.

Stirrup Hoe is just another form of bash the weed into the ground and cut it up. Bash it and pull it methods fail against particular classes of weeds. When bash it and pull it methods fail options are either chemicals that is not environmental friendly or heat methods(flame touches, steam jets and boiling water) that are not human friendly.

You will see flame touches put up as an heat methods these do have major set of downsides of being able to set garden alight as well as being able to burn human.

Chemical weed treatment comes out of attempting to find ways of killing weeds without starting fires or getting humans burnt. Robot provides as with another way to use heat methods without putting human in harms way.

Its always key to remember why particular things came into existence when you attempt not to use them.

The strictly written media is a horrible way to communicate, since we normally reply so much on seeing an individual’s body language, facial expressions, and tone of voice. Please understand I am NOT wanting to sound like I am rejecting what you are saying, or meaning to be demeaning.

I completely understand your points on obnoxious weeds that (on a manual weeding basis) seem to only submit to things like digging out, chemicals, or boiling water. However, it appears that you are still thinking in terms of a human weeding schedule. We have a lot of other activities that we do (work, eat, entertainment, etc.) We are not prepared to spend every hour of daylight weeding a garden plot.

That is the whole point of FarmBot. It CAN spend 12 (or more) hours a day taking care of the garden, and will not get tired or bored. Even weeds need sunlight to grow. They have limited reserves (particularly while young). If FarmBot does a weed check every 2 hours (or maybe just every 4 hours) and cuts off the portion above ground as soon as the weed is detected, they will never get well enough established to require the more extreme measures that you propose. Cutting the top off a plant at the one or two leaf stage is going to inflict terminal damage in most cases. They will simply run out of reserves (from the seed) and die.

A weed growing into the bed as an extension of a weed OUTSIDE the bed (quack grass) will be a different matter, and will require direct human interaction. This would be a good piece of code to add to the software: if a weed returns in the same location for more than (3-4 days? 1 week?) a specified period of time, an alert is sent to the owner. The owner can then direct WeedBot to position the head over the “offending” weeded area, and the owner can then manually intervene.

All I can say is woohoo… Take that Monsanto! Lol

a couple of things here…
I would want a no chemicals approach to weeding with Farmbot!

It is totally unnecessary to use chemicals!
With 24 hour survalience you can take them out anytime! No modification of the crop! No poison added to our food! Get them when they are young! I like the idea of pulling them and laying them in the sun when young!!! This may kill the bad ones like buttercup, but, you must remove all soil from the root!!! I usually wack them (on my knee or something) to fling off the soil then lay them next to a good plant to mulch it!!!.. Morning glory is another matter though! I find the center of it and pull it up and take it out of the garden! Not wanting to drop seeds!!!

When I weed now, i get down and dirty! I pick out the small ones with a pincher grip, and lay them in the sun… If it’s one straight young root like shaperds purse, this kills them within a 1/2 hr! (Even when older, but the seeds are there to complicate the progress)… (Better to get them before seeds are there!! or remove them and shred them!) sometimes I throw them on the lawn and then mow with the bagger to collect the organic material…a bit of composting with the grass clippings and it’s good mulch(which also helps keep the weeds down!

anything that gets roots in first, like buttercup, or quack grass, burdock and dandylion! I pry the roots out, (as much as I can get) put them in a bucket and remove from the garden… . I use a chipper to make the pieces small and after it is composted return it to the soil. Of course I don’t have the time to do it all like I want…that’s why I’m excited about the Farmbot!

leon.kemp there is a point of effectiveness. Bashing weeds every 2-4 hours is most likely over kill.

Does pay to read. This is general garden prep tricks.
4 inches All the seeds are dormant. Only bulbs and tubers will sprout and grow at this depth.

Weeds in that grow from seed in the top 1 to 2 inches will be beaten by bashing them quite simply. Weeds from the 4 inches or deeper that are bulb and tuber class weed chemicals/hot water. Nut grass is 8 to 14 inchs into the ground. Boiling hot water does down to 18 inchs. Issue with something like bulb and tuber class weeds like Nut grass a few early bashing will not kill it so could survive to the point where farmbot cannot detect it any more from the wanted plant. Bulb/tuber weed has quite a bit of reserves. So no the problem weed does not have to be coming from purely outside the bed. Failure to do proper prep to kill tuber/bulb class weeds when building a bed can bring on hell.

There is another class that are grow from every broken off segment class I call these “succulent class” of weed because most of them are every bit of it is basically a seed. Does not matter how many times you bash these they are not dying by that method. Yes bird drops a bit of succulent class weed on garden bed machine keeps on bashing it all it doing is making problem worse.

The reality with a weed that is going to die from direct bashing the number of times to bash it is about twice. Why twice is first time you might not have broken it well enough. Third time you would want to drop boiling water on it in because it most likely something more stubborn being tuber, blub or succulent class. Cone shape hot water makes when you drop it on a weed means you don’t want the wanted plants roots to be large enough to be in the hot waters strike zone. Yes kill the weed early is your objective.

This is another common mistake people make is weed more frequently this will increase weed kill rate. Reality it does not. All the plants include weeds from the top 1 to 2 inches once every 24 hours times 2 is max kill rate this has been studied for ploughing weeds in if the weed is living though that you are not looking at a weed from the top 1 to 2 inches but at a weed from the 4 inches and deeper class being tubers and bulbs or grow from every broken off segment class.

In fact most cases you can go slower than once ever 24 hours and still stay way ahead of the weeds. Once every 1 to 7 days was depending on local weed populations and time of year has been found quite successful in many areas. The key question is how long does it take weed to recover from putting on seed leaves to get to true leaves for the top 1 to 2 inches plants that process takes over a week this include what we call weeds. Plants including weeds are not exactly Speedy Gonzales in most cases. Where things go wrong with people who only garden on the weekend is they miss one weekend and now two weeks have passed weeds have been able to nicely recover.

leon.kemp basically there is quite a bit of science to this.

Quackgrass under ground growth rate is 2.5cm a day basically 1 inch and that is in ideal conditions. So in 3 days knowing that you have to deal with it the plant at that stage should be small enough to kill with hot water. So quackgrass area would be once every 24 hours to be on top of it in 3 days. If it takes you 14 days to address it then its a nightmare because you now have a 14 inch across plant to deal with that will not be going down easily.

This may kill the bad ones like buttercup, but, you must remove all soil from the root!!!
Or hit with boiling hot water. That kills buttercup, quack grass, burdock and dandylion with it still in the ground. Hot water and laying them in the sun is basically the same process cook them but there is a key difference.

Palomineo I have done a lot of no dig gardening were letting the dead weed remain in ground. There is something interesting about the hot water method compared to wasted all the effort digging the plant out. You give the weed the plant equal to human gangrene using hot water. Sections of the weed/plant you have killed under ground with hot water start rotting and that rotting goes on to the sections of roots you did not hit with the hot water. Prying the roots out turns out to be the hard way because the snapped of bits may avoid rot so plant gets to come back around for another pass. Of course plants with leaves above surface can fight off the rot but for best plant growth you wish to avoid this. Reason for attempting cutting off/bashing first is less disruption to plants around the weed. Digging weed up is quite disruptive.

From my point of view being able to drop hot water on a pest plant early saves a lot of hassles latter on. Bash twice if it comes back after that hit it with hot water job done.

24 hour surveillance is not exactly a use against weeds. Once per day surveillance is enough against weed. 3 to 4 time a day would be more critter control and fungus control.