Location of autonomous bot? Creating a bot to work a plot instead of gantry?

I realize there is a place for the gantry/Farmbot, but a mobile platform to work a plot also can be desirable for certain applications. I am considering designing one that can be used for forestry, but initially on a limited flat farm area…

I have a stable bot platform chosen, I am hoping to hear of what kind of bot location strategies are out there… Anybody got one? Gps could be the ultimate choice, but for a limited flat farm area would drive the cost up a little too steeply!

Have your looked at Ardupilot
Also Reach RTK

Hi Josh
Yes that would certainly help it to be mobile…I am looking for a way to securely and accurately get the location of the bot… It will be sitting on a variable plot. It can become a stable six legged or tripod supported platform, allowing scanning, photographing, locating, tool deployment, watering, ect!

I was thinking of perhaps logging a location using perhaps corner and height markers. Led or laser locating.
They could interact with the bot to Id a correct location and height within a few mm. Plant data, watering and weed removal could then be accomplished!
Thanks hey!

What do you mean by Variable plot. Are you thinking the bot will move from one plot to another?
Are you thinking the bot will still have a Cartesian style platform for deployment? The only other way I can think of to get accurate location is to have markers. As you stated LED Or beacons.
RTK will get you with in one centimeter. As you zero in on the location you will need beacons.
If you want to get more advanced you can use SLAM. This requires more processing power. But what you are talking about doing is very advanced. I am working on a mobile platform. I was planning on using the camera. It would work very similar to this. Locating fiducial.

Variable plot would be plot size and location…

I don’t want it to be limited to just one size or shape, or elevation even. The platform would also be able to track to a location… Set itself in the stable position and work from there… Sort of like the feducial does.

Move to the next location and pull the plant information …positions when the last watering was done, ect. and perhaps water them, plant something in an open spot, kill a weed… Scare a deer put of the plot! You’d want it to be versatile so it doesn’t waste a lot of time going from one place to another… … Non linear… Grab another tool and do what needs to be done there. Potential for more than one working the plot.

Maybe have a sequence to perform…water all in this Area… Is there any seeds to be planted, weeds to pour boiling water on.

Ardipilot is one of the most advanced platforms for navigation. This system will also need cellular to broadcast back to a base. Ardupilot uses radio to transmit telemetry back to the base. My plan is similar. I am not trying to change elevation to start with. I plan to us encoded wheels to move to a general location then use 3d printed stakes with disks to act as fiducial markers. A qr code could be placed on the disk to id the location maybe. Elivation is not addressed with this system. Maybe in the future but for now this is my project for the winter.

Ok second try, lol wrote this out once already! I’ll have to remember that I loose the draft if I leave the page!


Each leg on the platform uses a computer for control. I will be using a rasberry pi for this. Ethernet is included with the new rasberry pi ii or iii at about $35 each. Add a router or two and you’re networked for say a 100’ by 100’ area.

If I’m going with this already, why not continue… A rasberry pi for each corner unit finding location, running LEDs, locators, angles, lasers? ect. A pi for the more complicated tools? A cluster for fast heavy locating of info? Google has been working on this!
Hey… Here’s a 15 pi cluster in a homedepot toolbox… More than enough computing power and what about memory? The rasberry pi iii B has 1 gig ram, and an sd slot so 64 gig sd for each? 128g each? Hey!, I already have one of those!.. The toolbox I mean! Lol

Oops … 60 core… 15 pi … 16 G ram, close to 1T sd memory using 64G sd’s

Are you familiar with SLAM navigation? I wonder if this would work for your application? https://youtu.be/qpTS7kg9J3A.

Oh that does look useful! I have not seen it before! I will study it!

You could compare the original map with a newer one…marking the changes!

1 Like

With all these clustered computers, this could be termed a Beowulf cluster
And we’d have a Beowulf farmbot in our garden chasing the deer away!

It could happen!

1 Like

A Bot for handling wood in the final stages of heating house and Green house and Chicken coop into an outside boiler would be kind of neat. There is a lot of lift and lug at that point to keep a 500,000 BTU. boiler going. Run from a laptop along with the ability to turn on and off sensors, (heat, fans, water valves, pull a brush hog, and snow blow) might be fun!!! I have tried it.




Josh when you think about Elevation when dealing with plants most of your operational measurements are from real top of soil down. So Elevation of robot could be ignored a lot due to the fact tool depth into soil is the important bit. Location to know if you are looking at the same area or not is important…

Josh something to remember with SLAM is what will the robot do if it blinded.

Palomineo errors like above is why you cannot depend on GPS at times to get you to the right location on a good day.
Add in gps spoofing stuff you can say that GPS will never be 100 percent dependable. http://www.forbes.com/sites/parmyolson/2015/08/07/gps-spoofing-hackers-defcon/#245976cf7a48
Add in that gps spoofing is open source. https://github.com/osqzss/gps-sdr-sim
Add in pokemon go cheats using gps spoofing https://www.insinuator.net/2016/07/gotta-catch-em-all-worldwide-or-how-to-spoof-gps-to-cheat-at-pokemon-go/
Add in one fool who does not know how to rf shield there pokemon go cheating so now your poor machine has totally garbage GPS data…

I think we have to kind of lock SLAM or equal navigation no matter what because we cannot depend on any radio signal to be truthful.

1 Like

All great pionts. You can see in the linked video the camera was blinded when it was looking towards the sun.

Pokémon go gps hacks thats just funny. As far as hacking a tesla and taking control of car component. The need for better security in automotive industry is a well known. BMW had issues a few years ago. Hackers were able to start the cars through the diagnostic port. Making car theft easy. This is why I am working on a simple system. That will have limited range and speed. Also weight will be low. And no gps.

Add in that the government won’t let you have access to100% accurate gps(so you can’t use it for targeting missiles or snipering)

It would be best to consider gps for use as a general area tool and have a or perhaps 4 (corners of the plot) triangulation markers for more accurate plotting of the location. Height is rather important! Where do you place the watering tube? How tall should the platform be?, to be able to walk through the garden to weed where it was when it was so rudely interupted( by the deer) where is the soil surface when it’s planting? Is the weeding tool pushing down 3" or 3’3"

I have been away, taking care of things…work, life and other things have been keeping me away!

I have been considering how to make a bot that will not be exceptionally expensive, will be able to do all the the prototype Farmbot is able, and could be easily duplicated to expand the area of gardening!

I figure a 3’ tall six legged bot that can carry the Farmbot tools… The Odex design made from aluminum & powder coated. Carry water and all the tools stopping to replenish, when it’s empty. Program for a 10’ x 10 ’ area and build another one for each expansion of 10’ x 10’

Have a (6)set of rasberry pi, one for each leg and other sets for each of the tools. Have corner location and height pi and use the Bluetooth and wifi for communication

What have I forgotten?

I get that the goal being reached for is cooler b/c an fully automated bot is really cool, but would half measures be considered? If you have a moving farmbot setup that can handle any 1x2m area it is parked at then things get simpler dont they? Then you could design you garden as a series of 1x2m plots and the bot would only have to go to each one, know which one it is at, switch to the matching configuration for that plot.

If you do that then you could have your mobile bot, follow a line or rail to get there, use a signal strength to hone in, remember set movements, all sorts of options. Either way once it arrive it would lock into the plot, identify the plot by either set number or coded picture or other option, and then function as a farmbot until it needs to move on.

Hey there!
I’m not looking at doing it this way for it to be a cool robot… The idea is to make minimal changes to the area, and to be able to scale it up and down or even change the purpose

No rails or fixed pieces, means it could be easily modified… You could even pick it up, throw it in the car and take it to a new area, set up the corner markers and put it to work!

(Then you could design you garden as a series of 1x2m plots and the bot would only have to go to each one, know which one it is at, switch to the matching configuration for that plot.)
Yes! Or get another bot or two and have a 4x8 meter area garden and bots working it in sections!