The Newbs guide to Farmbots limiting factors and strengths

The Farm bot is unique in the farming world mainly because of its intense degree of physical Cartesian precision. This makes it ideal for planting seeds, spot watering, probably for transplanting, weeding, and eventually harvesting.

While in certain lights it suffers from scalability issues, ill come back to that later, but through the entire cycle of small primarily herbaceous plants, initial plantings and starters for trees bushes or shrubs; scalability comes in the form of a conveyor belt. IE human operator loads trays and planters with dirt, packs the operational size “supertray” until it’s full to spec… and then as planting and watering occur they unload the back bay.

Another major strength is also quite unique, it’s open source nature. This also makes the Farmbot very flexible and reduces development costs, I can think of ridiculous niche uses for a Farmbot, it could grow mushrooms, it could be used to grow coral and other sedentary marine life.

What can we really do with this? And what can we NOT do?

Let’s talk about potential design constraints. This will give us a long view idea of what additional development or optimization is necessary.

  • Scaling beyond small plants
  • Redundancy has to be eliminated. Having a 3d printer for every garden row will not scale up as well as a single unit that “docks” from row to row.
  • Pinpoint accuracy sometimes means “tons of time calibrating and leveling”
  • Weather

I definitely would appreciate any input here too :slight_smile:

Let’s scale up to massive size and eliminate redundancy; there’s three foreseeable possibilities…

  • “Little Red Wagon Bot” Or a robot segway not unlike Amazon’s warehouse packing robots. The bot could “dock” on the garden row, and click the gantry down onto the rails. After the row is planted, watered and/or harvested, the Farmbot returns to its resting position, and docks with the segway, and moves to its next task and bed.

  • Conveyor belt - while a conveyor belt can address scalability it comes with its own set of design constraints, plant height and grow bed weight, and cost. For planting, transplanting, harvesting and watering small starts or herbaceous plants it could be fantastic.

  • Build turnstile rails, it would look like a slow roller coaster for 3d printers. But it would snake around and connect in a giant infinity knot.

Don’t forget about the need for water, currently provided by a hose. Makes it hard to go “from row to row”. I think if you want to scale it up that high, you need boxes of soil on wheels. Then you roll them to the robot. The robot measures the orientation of the box before it starts. Also reads a barcode off the box, so it knows what plants are in the box.

2 Likes

I have been brainstorming this as well. I was looking at a rover style bot using ardupilot with Reach RTK. If it works on tractors why not a garden rover. Then I was sent a link to Farmbot. With the rectangular design of the bot this scales easily on the y axis. The x is limited. The mostly 3d printed CNC on thingiverse gave me the idea that the bot could move. The z axis would not need to be moved only, but the whole frame. When moved it would only need a z axis calibration. Imagine a Synchro drive with a slip ring for water and power supply. The frame would locate alignment anchors.

Limitations: the legs need to rotate. Legs may need extra support. Larger diameter tubing may be all it takes.

2 Likes

I think the easiest way to scale FarmBot to allow it to manage more plants at once is to make the tracks longer. For very little extra hardware the device could be 2, 4, 10 times as long. Probably beyond 2 or 3 times longer the x-axis motor would need to be upgraded to be able to drag all the extra tubing/wiring/drag chain,

The next easiest way to scale is by widening the gantry, probably up to twice as wide. Even wider would be possible if you had additional tracks and gantry columns to support the middle. This would also require an upgraded x-axis motor to move the extra weight.

4 Likes

It looks to me like you could get 4x the area and thus produce for perhaps 10% more cost. But I’m making a lot of assumptions that might be wrong.

  1. You can make the length longer by running more beams end-to-end along the sides of the planter, and making hoses and cables longer. Might need to rig a suspension for the cables/hoses to keep them from getting tangled in the plants.

  2. You can make the width wider by attaching multiple beams end-to-end. I’m not sure how the carrier rides on the beam, though - would the bolts and plates interfere with the carrier’s ability to run the whole length of the beam?

I don’t think taller would be important, since I’m not growing corn! :smiley: So no need to get a longer worm gear.

My concern is that there might be an engineering reason that Genesis is 1.5m x 3m. With 3D printers and CNC mills, large sizes are challenging because of wobble, but I wouldn’t think that a little wobble would be an issue with plants - even if it’s off by a full inch, that’s not a big deal. Right?

By the math, squares are most efficient - 6x6 meters is 8x the area of 1.5x3 meters, for perhaps a 10-20% increase in cost, which seems like a great deal, if you have the space. And it’d make the economics awesome!

Any thoughts or advice?

2 Likes

I’ve been reading through this section and thinking about possible use scenarios and had a thought.
what about a circular track pivoting around a central distribution hub. scale could be by both increasing the distance between rails and by increase in circle diameter. This does pose some programmatic challenges as opposed to a rectangular grid but similar too the large scale farming using a circular irrigation perhaps at some scale the idea may be workable. The other issue is zeroing the z axis - for something larger scale and for more simple initial set up perhaps some type of range finding device could be employed. perhaps optical or maybe a pressure probe attachment.

2 Likes

Great thanks thats definitely critical.

This idea would totally work albeit with one caveat; 100 gallons of wet dirt is very heavy (550-850lbs / ish), and has the same amount of space for roots as approximately 1-2 gallons of water (8-16lbs / exactly). I think/feel like if you move around your grow beds using water / media matrix substrate; you’ll get more efficiency. I know you said “soil” not “dirt” but either way in a gallon of most soil; there’s only .01 gallons of space left for roots.

Wow Josh this is exactly what I need to get me thinking. Im not as familiar with the electronics component of what you’re saying I’d love to get deeply into that I msged you with my FB info for msging. 100% agree legs will need to be of a unique design or just seriously overbuilt. I think simply overbuilding an existing design is sufficient.

In order for us to get out of the 4 ft by 12 ft, walled garden foot by foot mindset - think of bots wandering in our ecosystem and maintaining it with nothing but solar power. Not only that but, a ground-based mobile autonomous “farm” bot is going to dominate the commercial sector.

One idea would be a 2-wheel segway with an arm that can stake itself into the ground at least up to 6 ft away. That way it’s balance and positioning can be maintained while it moves from place to place. I was thinking a triangular shaped craft where the wheels not only rotate but the legs can also vary in length for positioning; but that’s not for the Little Red Wagon Bot - more for an all in one bot that defeats the need for tracks.

So when you say 2 or 3 times longer is that 20-30 ft you’re estimating before a x-axis motor upgrade would be required?

So I have a major question that could affect the limiting factors; for you Rory - without doing much R&D, using currently available products, and with motor and print head optimizations/tweaks/upgrades … how many square feet a day will one print head be able to steward? You could always add extra print heads but of course that changes $$$ bottom lines.

#1.) Yup! Its pretty common in industrial agriculture to build all sorts of riggings and suspensions. Also for riggings and suspensions, there will be potentially a maximum length anyways which could be defeated by adding power/water “stations” and quick release/snap systems.

#2.) I think Rory was technically accurate you’re only going to double the width easily without additional R&D but yes doubling it shouldn’t be too difficult. =-)

#3.) Taller is probably one of the best things to be thinking about. Think vertical, and stacking.

#4.) @roryaronson Care to elaborate on WOBBLE sir rory?

#.5) For watering an inch makes zero difference. For trying to remove a tiny tiny weed its pretty important, let alone zapping a moving aphid out of the air before it can lay eggs! and I suspect part of the strength of the FarmBot lies in its extreme precision

Why is everyone in this thread so cool and smart??? I typically despise hobnobbing with the “trolls” but why aren’t they in here?

This is a fantastic idea!!! you just gave me another one!!!

ok so you have one straight row that leads out from the central ring and water/power “station” which is in the middle right? that straight row connects (arbitrarily) lets say three concentric beds that are rings which are only broken in 1 place (where the straight row connects the railing in the middle so the print head can switch between the three beds.) Are we getting there?

@roryaronson Does a 4 ft wide ringlike circle grow bed present a programming challenge because the printer angle changes vector as the track curves?

Yes optical =-) then you just paint a little yellow dot at 0, 0, 0 and boooooom

Okay instead of rings… between your musing and this picture of a permaculture herb spiral … a new idea is forming which is circular and makes extremely efficient use of space - a spiraling track would prevent any need to exchange or relocate print heads - can be extended indefinitely, doesn’t need extra tressing to widen the bed… hmmmmm thoughts anyone?

While rectangular is possibly more practical for most people, I agree that a circular bed would be most efficient in terms of covering the most plantable area, particularly if a little wobble is acceptable. One challenge with circular robots that I see is that cartesian (X/Y/Z axis, rectangular) robots are very popular (from 3D printing and CNC milling) with plenty of firmware and software support, etc. That being said, the math of converting X/Y/Z to angle/distance/Z is pretty easy. The other thing that jumps out at me is that if the “arm” is only attached at the center it’ll tend to wobble around. And to fix that you’d need a circular outer track, which isn’t something that you can buy as easily as linear rail. So while a central pivot and arm approach works fine for (for example) spraying crops because you cover an area and don’t need to aim at anything, but it likely wouldn’t work for plucking weeds, etc.

Though come to think of it, a camera-based approach might work. That is, even though the precision of the arm wouldn’t be great, if the camera can be used to determine exactly where the ‘head’ is relative to where it wants to be, that’d allow it to compensate and hit the target precisely.

Fun to think about. This is all more complex than cartesian, …

1 Like

Hey everyone, thanks for chiming in here! Sorry I’m not as fast at responding (also have hundreds of emails to respond to, too!!)

So, regarding wobble - everyone here is pretty much right. Wobble does not have to be as controlled as with a CNC router or 3D printer, but it does need to be minimized for some precision operations like picking up and putting away the tools, and for taking photos. The main issue with scaling and wobble is that if you go twice as big (in certain directions), you’ll get something like four times or more of the wobble. So going taller would be difficult to do without significant strengthening of the hardware to make it more rigid. This of course adds cost and weight. Whereas making the tracks longer doesn’t increase wobble at all. So the best place to scale first is with the tracks length. Then the gantry width. Then the z-axis, and only if you want to grow corn or sunflowers… or fruit trees!

Polar coordinate bots would be awesome, and the software should support that with a simple firmware conversion from cartesian to polar coordinates. You could either have a circular “track” made of wood or bricks or compacted soil that would be enough to support a rubber wheel at the outer end of the “arm”. The support only has to be in the z-axis, ie it doesn’t need to be precisely attached to the “track” in the radial or tangential directions. The other option is a cantilevered arm that is supported by guide wires from above, like a suspension bridge. It could be counter-balanced with an arm sticking out the other end so that the moment force on the main vertical pole would be minimized. The circular track system could probably scale larger, because you could have multiple track rings. While the cantilevered system would be easier to setup for smaller devices.

Regarding scaling the cartesian bots, everyone here is correct that for not a whole lot more cost, you can get a lot more value from the device. In fact, we’ve already been working on FarmBot Genesis “XL” for a while, which has a double-width gantry (3m) and double length tracks (6m). You can see a sneak preview at the end of this video: https://youtu.be/gDocKosjY0I?t=3m50s

We weren’t planning to release larger devices initially because we wanted to focus our campaign on one product. But, with the latest hardware (which is beefier and more battle-ready) we’re learning that we’re pretty much ready to make the larger kits available for pre-order, and we’ve had SO much interest in them (see: this thread!). The larger devices will be the same height, but be expandable up to 12m in length, and 3m wide. The main cost increase will come from the extra extrusions, plates, and plastic components, and the longer tubing/wiring/cable carrier required. But, with current estimation we can bring the cost down from over $600/m^2 of growing area for the smallest size to under $150/m^2 at the largest size! Stay tuned to our newsletter for the coming announcement!

Also, maybe you all can help out: Right now there are two main issues we’re working out with the larger devices:

  1. By scaling in either the X or Y direction, the gantry becomes heavier because of the extra gantry weight and/or the extra tubing, cable carrier, and wiring it has to pull along the tracks. How can we move the gantry with this extra weight? Some options we’re exploring: Using a gearbox to increase the torque but decrease the speed of the x-axis motor. Adding a second x-axis motor to the other side of the gantry and using a second stepper driver to power it. Upgrading the x-axis motor to a NEMA 23, and powering it with a more powerful, independent stepper driver from the RAMPS shield. A combination of the above. We’re leaning towards using a gear reduction or dual motors because upgrading the driver electronics would be more expen$ive. Further, we’re leaning towards dual motors because then we can maintain higher speeds and eliminate the driveshaft!
  2. Maintaining vacuum pressure with such a long tube for the vacuum line (for the seed injector). We haven’t done enough testing yet. But three obvious solutions are: get a more powerful vacuum pump, increase the tube diameter to reduce pressure losses, move the vacuum pump closer to the UTM, such as onto the gantry instead of next to the bot.

Cheers to open-source!

2 Likes

Just found this product, site, etc and got to this forum for a once over. It’s exciting to see what’s happening here!

jeabrams beat me to the idea of a circular setup. Don’t know if it’s the answer but definitely believe with motivation and innovation this could be a fundamental change to the way we grow plants on large and small scales.

Maintaining it and logistics would be two areas to address as well which have been commented on.

Great stuff guys and gals!

2 Likes

Brand new here, and am actively studying the Farmbot.

I’m am particularly interested in the ability to scale this at row length levels. I’m a small sustainable/organic farmer, and we tend to use very specific row widths/lengths that are based around the small scale equipment we use (generally 30" to 60" row widths, 50 or 100 feet in length).

The same problems that Farmbot solves in a backyard scenario are exactly the major hurdles that we face in production small scale agriculture (efficient seeding/spacing, watering and most importantly, weed control). If I had something like this available to me I could easily double or even quadruple my production while cutting labor in half. This could easily be a game changer for small farmers.

I’m off to go learn more, but I would love to discuss and possibly help with experimenting at scaling this at a small scale farm level!

Cheers!
Jody

5 Likes

“particularly if a little wobble is acceptable.” - sounds legit wobble shouldn’t be an issue

"One challenge with circular robots that I see is that cartesian (X/Y/Z axis, rectangular) robots are very popular (from 3D printing and CNC milling) with plenty of firmware and software support, etc. That being said, the math of converting X/Y/Z to angle/distance/Z is pretty easy. " - I think @roryaronson would be more knowledgeable here - sounds like swapping over to a polar coordinate system might be easiest. I can think of some novel ways to integrate a kinect sensor, some calibrations and a cartesian system to operate a spiral or circular space, but that would complicate setup/require scripting.

“The other thing that jumps out at me is that if the “arm” is only attached at the center it’ll tend to wobble around.” - Im not 100% on this @jeabrams idea originally - I believe its just the power/water that comes from the center, the print head is mounted on the motors and track.

“Though come to think of it, a camera-based approach might work. That is, even though the precision of the arm wouldn’t be great, if the camera can be used to determine exactly where the ‘head’ is relative to where it wants to be, that’d allow it to compensate and hit the target precisely.” - Yes, I believe theres a number of camera oriented functions already happening. If you dig in a little deeper you’ll find some people trying to design a pest deterring laser that can spot kill bad bugs and bite critters.