Farmbot for small businesses or people who want to do a little more. Why extend tracks when you can put to existing genesis or genesis XL a set of wheels to replicate the squares a couple of times.
If you put a water pipe and a few plugs along seems that can be done. What do you say?
There are a lot of changes for this to happen but I’m doing it with a bunch of guys, a fully equipped farmbot on wheels, open source
We’ll have nice things to show this month, so we’ll post some photos and comments in the next weeks
@mcx, I think that putting it on wheels will create a lot of challenges, because it will be difficult for the farmbot to know its location. This could be solved though, with cameras and positioning systems and such, but quite challenging. I think a good place to start is to put it on rails, I’m working on such a project here. But maybe rails is an unnecessary intermediate step, maybe we should focus on creating one on wheels directly? What do you think?
it isn’t cheap, however very accurate gps is readily available: https://www.swiftnav.com/piksi-multi they claim centimeter level accuracy which would likely be enough. They are not cheap @ $2.5k, however adding wheels will clearly increase the value of the farmbot.
At first I do not think it needs to be with gps. To replicate 3 squares in line for ex. you need the farmbot to recognize each separatly and set it from software to go forward or backward from 1 to 3 (3 to 1) the amount of distance you need, same every time.
If wheels are used, some form of feedback will be necessary so that it can correct for errors and minor deviations. 1 example would be if there is mud or wet leaves under the wheels, one wheel might need to run for slightly longer than the others, or the robot might need to turn slightly to correct for a clump of mud or uneven ground under a wheel.This would not necessarily need to be gps, but that would likely be a more simple solution than trying to use vision and/or
accelerometers for a similar purpose.
Pictures will come please be patient, we want to amaze everybuddies with good ones but we’re not yet ready. At the moment we’re implementing electricity things : batteries, solar panel and wind turbine !
Our view about the GPS is that we’re not going to use it at first, we are focusing on using human remote control to go from one parcel (plot) to another. To identify and adjust the position according to the farmdesign, we’ll use small planted pickets with a QR-code on it. So that when the farmbot is on site, it can adjust (opencv) and load the allocated information (farmdesign, sequences import), and the farmer when on site can scan the QR-code to get information (past actions, measures, calendar, …).
Our opinion is that the GPS/Galileo will be best used at low cost for low-precision things, like moving to refill the tank (then adjust water pipe with QR-codes) or moving from one parcel to another or go back to the farm.
Just imagine this story : You’ve prepared 10 parcels (two parcels can have same farmdesign), 1 square meter each. Grab 10 pickets, stick the QR-codes, go to the meadow, plant each picket where you want each farmdesign to be applied, and you scan the QR-codes with your smartphone. We’ll have GPS location from the smartphone. Then you release the farmbot into the meadow. Or you can ask the farmbot to choose spots and plant the pickets by itself.
That’s what we’re thinking about, do you have some suggestions ?
Using QR-codes to identify parcels is a must. From previous answer we can see where you are going.
Even if you use human remote control it is very usefull to make a software module where we can make paths. In the future if I let Farmbot on a meadow I do not want it to go how he wants to a specific point.
Lets focus first on double the area. You can apply Moore law for a couple of years.
So for example If we put 2 framebeds in line if we can program farmbot (on wheels) to go a few meters back and forward with QR-code calibration( maybe add some more optic endstops like the 3d printer ones), and the option to set the distance between the framebeds will do just fine. The frame beds linked togheter will not let the farmbot to slip and with endstop he will know exactly where he is.
I like where you are going , but the rate of accepting farmbot by more people is crucial and that depends on a bigger area to make it more economicaly feasible, and in order to amaze everybody please do not make huge steps that take a lot of time, that need a lot of testing to make it stable.
If I can help from distance please let me know.
Interesting topic, I guess it’s hard to do all the tasks the farmbot is dowing and make it mobile, hard to bring water for example for a large areal but weeding and placing seeds I guess is doable.
I think that the camera would be enough to detect wheel slip or an accelerometer if camera isn’t enough and a couple of qr codes, pinned to the ground, on the field to re-calibrate.
A bit like this; https://www.youtube.com/watch?v=swcu6KxL04E
The reason why I was assuming high-resolution gps would be easier is that you can literally read xyz coordinates from the device. With qr codes, there will be a bit more software in order to collect data from a camera, pass it into open cv, filter the results from open cv, plant stakes in the ground, use a phone to capture gps position of stakes, send data from phone to robot, etc. At this point I am lazy and figure there’ll be enough other problems that paying for a high resolution gps to avoid some complexity would be nice. I understand if you have a different perspective on this though.
I’m definitely interested in seeing the progress you make whenever you get a chance. Thank you for sharing.
yes maybe gps is a good idea, but i guess it needs some help to get precise enough to pick weed?
I found this project today; https://hackaday.io/project/21744-weedinator#menu-description
whick looks quite promising, and is even refering to farmbot. even though Id prefer much simpler wheel and suspension arrangement. Couldn’t find the user or any info about weedinator on this forum. My thought is to use 2 electric bicycle wheels 20"(as i have a few lying around), to take the most load, probably with something wider in metal instead of tires, and 2 smaller swiveling wheels to support.
i dont get why weedinator is still using a 3rd axis when the whole thing is on wheels.
and definatley a small battery and mechanical weedpicker instead of combustion engine and flame thrower.
good find. I’ve also been thinking about using bicycle wheels.
GPS that is in your phone is only accurate to a meter or so, however there are high precision gps systems which can get to 2 centimeter accuracy. This might not be accurate enough for every task, but I imagine it should be accurate enough for most tasks.
The wheel in the pic is a 406mm bike wheel but not motorized, my idea is to use one screw and nut for each profile piece and drill holes through the rim, maybe do a little grinding to make it harder for the pieces to turn.
Found a stroller with 2 swiveling wheels the other day in the trench, but was gone when I went back to pick it up yesterday.
Gordon Moore appeared on stage in 2007 and said his named law would no longer be valid in 10 to 15 years.
That’s 2018 and he’s right the farmbot on wheel can address 200 squared meters every day
Which is beyond expectations according to the fact a very few percent of parcels will need a visit every day. We’re enthusiast with Rory’s statement about the 24 volts migration in the next batch, it can at least double the perimeter by fast moving the 3 axis and strenghtening their positions while they are working on a slope for example.
I understand about the aligned beds it seems a more simple way with some common problems! It could be a easy milestone between human and autonomous control.
We think that randoming the farmbed locations will promote bio-diversity and make smarter usage of the soil regeneration mechanisms, so we’re also thinking to use it in wilder areas too and aligned beds will be rare. It needs to be autonomous about the paths at long terms, mainly because it will also needs cooperation systems, shortest path behavior etc
Yes very accurate GPS would be so fine but our contribution is not what you said @zdwiel : at first you will plant pickets with a QR-code, and the picket is displayed in the farmdesign so the robot will first use the GPS and bring the perfection according to QR-code position on the farmdesign (one farmdesign can have multiple pickets for better positioning).
As I understand, one GPS alone is not enough (2cm min is not enough for precise agriculture) so you might need to start with differential GPS and the second one should not be on the same vehicle so you need an other fixed point in the field (not so good for mobility), I think you also need the vehicle’s direction in space (x-y) and the Z position for the ground distance etc.
For the Z position, it needs to adapt for few centimeters but it can be important when seeding for example. For this we have Time of Flight sensors which will permit to measure distances.
For the direction, the QR code will permit this. Otherwise with GPS you might prefer to use an electronic compass.
I’m sorry @mcx, we need to make one single huge step in private at first because we’ve thought of a way of thinking agriculture in the modern world more than a tech solution and it needs a complete functional architecture we can rely on for few years (at best) and build economic vectors to support the effort and tomorrow’s openness. We also don’t want to promote any of our products or brands here, just talk about opinions and tech things with a great community.
I am definitely curious about high precision qr code positioning. Do you know of any good resources on this topic?
I agree that gps alone would not be enough, a compass/gyrometer/accelerometer would be necessary for determining orientation. Are the time-of-flight sensors you are using for measuring distance to the soil?
I suggest to read the getcornerpixel function in this code, x/y of each can help determine the line between pattern to see if farmbot is in line and try fixing orientation (thanks for the word
Usage of TOF sensors : to follow growth and want to contribute to the openfarm.cc project with those data (and photos) and the weeder could also benefit from this by improving detection mechanisms just a bit.
May be someone can help work on the openfarm.cc side to accept the plant data ( cloud of points : volume, size + photos) and help determine if there is a need now for an aggregating system between time elapsed since germination and size/volume ? I’ll ask on their slack