Penetrating dense upper foliage of intensive garden

I hope this is not a duplicate, but I did do my due diligence searching. I also scanned the whitepaper, documentation, and watched several videos. I’m very seriously considering a Farmbot kit, but so far have one major question.

Once the plants have been growing for awhile and the foliage has filled in, I’m wondering how the Farmbot works. I know in my raised bed this year, from above I could not see the soil through my collard greens or broccoli or Brussels sprouts.

So I’m not sure how the Farmbot vision system could identify weeds in this case; perhaps it lowers down through the top foliage nearer the soil and scans from there, and it knows it can push the UTM through leaves and small stems because it knows where the plants are rooted and therefore where the foliage is penetrable?

Also, it appears from the videos that Farmbot waters from above the foliage. With its minimized amount of water, accuracy is critical, and I’d fear that water from above would result in the water running off the leaves of some plants and not hit the roots. If the UTM were lowered through the foliage, this would be much less of an issue.

Any comments would be very much appreciated.

FarmBot is not well suited to getting beneath dense upper foliage of a garden. Moving the z-axis down through the foliage may result in damage to the plants (much different than finessing a human hand though the foliage), and a retracting z-axis could get stuck on a thicker branch/stem and cause the FarmBot to get stuck.

The vision and weeding system is meant for early detection and removal of small weeds (read more here) and is best suited to combat weeds in the early stages of a garden in order to give your plants a significant head start. It is not designed to detect or remove weeds from a dense, late stage garden.

Regarding watering, you are correct that some amount of water will run off the leaves and potentially never make it to the roots. You could try to lower the UTM through the foliage, but it would really depend on the plant type and density to determine if you would be successful in this or not. Another option could be developing a side-spray type watering nozzle that could maybe allow you to lower the UTM close to the ground and then spray water sideways to the stem of the plant. That’s just an idea though and would most likely require a few prototypes to get it right, if its even feasible.

Thank you very much for the response and ideas. I’ll noodle on it awhile.

Hi @filo,

I don’t know the answers to your questions, but I would very much like to hear some more questions you have!

The reason for this is that I’m working on a tweak/improvement in the FarmBot eco system. You can find more information about this work here:

As part of this work I would like to “pick the brain” of real or potential FarmBot users. Would you be willing to answer some questions on a Skype call?

Hey filo,

from my experience so far, as soon as a plant is that big, it does not need to compete with weeds because it is already stronger than the weeds are. The water will rain from the leave but it will somehow get to the roots, otherwise all plants with big leafes would not survive outdoors :wink: :stuck_out_tongue:

Indeed it depends on the weeds and on the arrangement of the plants. Thus there will be no way around that the bot takes the age and growth of the neighbouring plants into account. All of these scenarios could be “solved” with neural nets in future, which would be such a nice science project to develop…

And please note also that there is no fully automatically working FB so far to my knowledge, please correct me if I am wrong… the project itself is still in alpha to beta state…


I am genuinely interested to know the answer to this question :
Is there a fully automatically working Farmbot ?

Maybe the one in rory’s backyard ?

@jerome please define fully automatically working?
You need to program the bot yourself, maybe one is able to program the soil moisture sensor in such a way that the bot will only water if the soil is dry, but I guess thats all. The whole system is too unstable to really let it automatically run over more than maybe 2 weeks or so…

But yes, it would be very interesting to get an answer!!! I think it will be no, unfortunately…

You got it : Once you had your sequences & regimen programmed, how sure can you get
that each and every sequence/regimen are performed accordingly to your program ?

In my experience with genesis 1.3, sometimes FB will stop in the middle of sequence with no
way to understand why that’s happened.

That’s why i’m not so confident for unattended setup. As an educational/research plateform FB
is great. As a professional tool to grow vegetable not quite yet.

But that’s also an area where the FB community could make a difference, eventually.

That’s why I asked. I guess the repeatability of the tasks once they are programmed correctly should not be the problem. Its rather the problem that there a movement issues, unaccuracies while moving and picking up tools, no prohibition of the arm to drive somewhere it shouldn´t, because its for example one mm off the position etc… and of course the reliability of the web-app and the background server applications…

What’s discussed in that’s thread could be an answer :

Also thinking about it, a new sequence instruction like “get_current_position” could allow to
test/compensate for movement programmatically.

Is there another way from within a sequence to test if a movement request has been correctly performed ?

Not to my knowledge, there´s only the option to be notified via log message or e-mail currently…?!