The world of automated farming may be an unglamorous one to those not invested in its attractions, but like the robots themselves that quietly get on in the background with tending crops, those who follow that path spend many seasons refining their designs. The Acorn is a newly-open-sourced robot from Twisted Fields, a Californian research farm, and it provides a fascinating look at the progress of a farming robot design from germination onwards.
The Acorn is not a CNC gantry for small intensive gardens in the manner of designs such as the Farmbot, instead it’s an autonomous solar-powered rover intended for larger farms which will cruise the fields continuously tending to the plants in its patch. It’s a work in progress, so what we see is the completed rover with the tools and machine vision to follow. It pursues the course of a low-cost lightweight platform, an aluminium chassis surmounted by the solar panel, with mountain bike front fork derived wheels at each corner. It has four wheel drive and four wheel steering, meaning that it can traverse the roughest of farmland. We can see its progress since a 2019 prototype, and while it seems as slow as the seasons themselves to mature, we can see that the final version could be a significantly useful machine on a small farm.
It’s not the first autonomous farming robot we’ve seen over the years, as for example this slightly more robust Australian model. We’re guessing that this is the direction autonomous farming is likely to take, with the more traditional tractor-based machinery projected by some manufacturers taking on repetitive loading and hauling roles.
Lightweight if somewhat noisy platform.
Looks so lightweight, it’s unclear if the platform would be capable of carrying a payload that can do actual work. And if so, for what duration.
Noise, that is a feature. (Scarecrow)
Yes.. Scarecrow feature, love it! Also, I assume this could carry a cooler of beer, the noise will alert me to when to turn around and grab a cold one while I am weeding the hopyard.
Yep we will be demonstrating tool use in the future. We wanted to start sharing what we’re working on and build a following of interested folks.
The noise always bothered me but I’ve just replaced the drive system with some geared hub motors which are very quiet! I’ll post a new video about it soon. :)
I estimate it weights as much as 4 Wal-Mart (heavy) mountain bikes. Mountain bike wheels can hold hundreds of pounds, and it’s frame looks solid enough to hold several hundred pounds of additional gear.
it a really good idea that’s doomed because if you put about a £1000 worth of aluminium, lipos and solar pannel, motors etc in a field in a form factor that will fit in the back of a transit van some b£$%!”$d will come and put it in the back of their transit van.
dont tell them till I grab a few first
Equip it with a very power laser and teach it to be a terminator. Pew-pew!
It doesn’t appear to do anything except drive in a circle.. Also it looks quite fragile and just wonder exactly what sort of work can it do. I don’t see any appendages on it.
Appendages are coming! As the article states it is still under development. We finished the base vehicle and now are working on a vision system and tools.
Yeah – nice and all but – what does it actually do, besides driving around?
As the blog says, it is a test platform. CV and implements will come later.
Acorn is a prototype that is still under development. This announcement is to let people know we have just open sourced what we have so far, which is the autonomous GPS guided solar powered vehicle and its associated Python code, PCB designs, and mechanicals.
We’ve just added a vision system to it which will be under development for a while, and next we’ll be developing smart tools. We want to use it for weeding, planting, pruning damaged plants, harvesting, and more.
Part of the reason we have open sourced it is to try to build a community of followers. We know some people will just like the idea but others will be farmers with real experience, and it would be great to get their feedback on tool ideas. We’re already working with local farmers, but more collaboration doesn’t hurt!
It’s to drive people to donate and help build a YouTube channel to also donate money into the pockets of some people that built a useless thing that runs over weeds or rows of weeds in there backyard or most likely someone else’s who doesn’t know.
“We’re guessing that this is the direction autonomous farming is likely to take, with the more traditional tractor-based machinery projected by some manufacturers taking on repetitive loading and hauling roles.”
Might have to build walls to keep people from coming in and taking their jobs. :-p
I didn’t see it do anything that would be useful to a farmer.
Kids can push a cart around a field all day, and they’re solar-powered, via the food we grow!
It’s true! Right now acorn does not do anything useful to a farmer, as it is still under development. We’re building robotic tools that will fit underneath the frame and work the field. But first we needed a vehicle capable of bringing tools to each plant. We’re quite direct about the unfinished nature of the project in our linked announcement post. But we’re ready to start showing people what we have in part to collect ideas on which tools would be best!
I have been farming wild blueberries for many years. They are a permanent cover and there is no plowing, etc. involved. Weeding is done with chemicals, but a robot, or fleet of robots, which roved over the field on narrow wheels 24/7 could potentially pick an awful lot of weeds and eliminate the need for some of these chemicals.
A wild blueberry field is also a patchwork of different clones which have different flavour profiles and ripen at different times. Right now, they’re all just scooped up over a period of a month and sent to a freezer plant. However, a robot which could distinguish different clones and gently pick the fruit into packages would allow a good portion of the crop to sold at the farm gate for triple the price I receive now.
A final thought is that I’m over sixty now and would like to avoid some of the heavier, repetitive work of harvest, but I could still use the income. A robot working ten acres (4 hectares) intensively could address that. I have been watching and waiting for a robot like this for over ten years because I can see the promise, but lack the know-how to create it. I wish you continued success in your efforts.
Hi Robert, I work with a company that is currently developing an autonomous farming robot in Australia. We have a sturdy base which can navigate farmland with ease, carry heavy loads including a chemical sprayer. I was just wondering if have any tasks which you would find especially useful from a robot such as this?
Computer vision is a development area of ours right now to identify when crops are ripe alongside picking of them. We are also adding in environmental sensors such as humidity temperature light and more to provide automated information to a farmer about his farm.
Hi Jared,
Thanks for your email. A team at Dalhousie University in Nova Scotia has been working on a “smart” sprayer for a number of years. I think the challenge has been around computer vision since there are so many different kinds of plants in a wild blueberry field. Wild blueberries, by the way, are smaller than the high bush ones you may be more used to. They’re about the size of pea. Anyway, the Dalhousie team has been working on a boom sprayer with independant nozzles which can spot spray weeds as it crosses a field. This would be very useful for farmers and reduce costs.
I think farming wild blueberries organically would take a big leap forward if tree seedlings and plants such as St. John’s wort, bulrush and so on could be pulled by a robot. Organic blueberries fetch double the price of conventional fruit, but it’s really hard to manage when you can’t plow, harrow, etc.
Wild blueberry fields consist of many (I’ve heard 47) different clones which occur in patches. These clones may ripen at slightly different times and grow to (slightly) different heights. A system of mapping the field might allow a farmer to tailor fertilizer to the particular clone and to monitor the yield of various clones and perhaps identify patterns of interest.
Another promising area is harvesting. A typical field around here (eastern Canada) yields about 3,500lbs of wild blueberries per acre. Harvest happens over a period of roughly 4 weeks. The berries are harvested with a tractor operated harvester. This harvester looks a bit like a rotating barrel three feet wide with 12 – 16 rows of steel rakes on it. The fruit fall onto a conveyor belt inside the barrel and are carried out onto another conveyor and from there into a bin which holds about 250 lbs of fruit. This bin is taken to a processing plant off-farm where the fruit is sorted, cleaned and individually quick frozen.
I wonder if a robot capable of picking blueberries cleanly could allow a farmer to offer blueberries for sale directly from the farm. If a robot could mimic hand-picking, it might even be possible to freeze the berries in packages of ten pounds or so without the need for IQF technology and yet still have the fruit emerge from the freezer mostly as individual blueberries instead of a frozen lump. – Rob
It can detect when crops have grown higher than it’s clearance, by burning the motors out trying to move.
How did you know? That’s inside information
https://allpoetry.com/All-Watched-Over-By-Machines-Of-Loving-Grace
So excited to see this here, thank you for the write up!
What we want most right now is followers, so please check out our YouTube video and subscribe to our channel. Thank you!
https://youtu.be/NsyEIgKVM5E
Hey Taylor just a little question? Why do you have the solar panels on the farmot itself? It seems to me that some more batteries on the farmbot and a central charging station, that can be solar powered or grid powered during cloudy days, would be more efficient. Also you could have multiple units recharge at the charging station, further reducing the cost of solar cells.
Thanks for the question. Well, we really like the idea of running directly off of solar power. Right now acorn does not have any batteries, just some super capacitors to smooth out power from the sun.
The advantage of this system is that it is fully self contained. No need to ever plug it in. That said, I do have a battery for it that I use for testing at night since I’m a night owl. It would be easy for us to offer a purely battery powered version that supported external charging. But I do like that this robot works every day as long as the sun is shining. We’re still experimenting with form factor and the good news is it’s really easy to change the power system.
Well for thousands of years farmers worked from sunrise to sunset, no reason robots shouldn’t do the same. :)
It could be both solar powered for strictly daytime use AND have a battery for heavy use days or nighttime. For example, 80% of the time it may rely almost solely on sunlight, while during planting or cultivation periods, it might rely on the battery for 60% of its power needs, returning in the evening for a wireless recharge.
Not designed by an engineer: 4 drive motors AND 4 steering motors! Efficiency coefficient < 0.125. Good idea overall though.
Could you explain this more? Is 4 steering motors that big of a deal (they are used just a tiny fraction compared to drive motors, and arent putting out a ton of torque as long as things are moving forward and its not making sharp turns with the tires up against rigid mounds). But how did you come up with efficiency coefficient < 0.125? Doubling the number of drive motors doesnt make it half as efficient….
Rusty hydrogen isn’t referring to electrical efficiency. “Design efficiency” is a measure of how well a design has been simplified and unnecessary parts removed. It’s a concept used in DFMA a lot, popularised by Boothroyd (Google it, it’s a great book for people designing commercial products).
The basic premise is that every part is a potential source of failure and an added cost. When designing for volume manufacture it’s almost always cheaper to make one complex part than assemble multiple simpler ones so good design seeks to reduce these.
I’d argue that the objection doesn’t apply here because the project isn’t trying to be for volume manufacture and modularity and development speed are more important for a project like this.
However, I agree with them that 8 actuators does seem excessive to achieve control over 2 degrees of freedom (forward-back and turning)or possibly 3 if you want to be able to translate left-right without turning.
Frankly though, who cares? It’s a great project and if they get the hard bit (CV controlled tool use) working it’s not going to be hard for someone to contribute mechanical improvements later on.
I will start by disagreeing, then end up agreeing at least partially.
Perhaps Design for Maintainability (DfM) is a more important issue here? While a reduced component count may help with manufacturing costs and does help eliminate failure points, a “farmbot” type device (or any long-life field operated system) is going to experience failures that will require remote maintenance. Since the parts being discussed are drive train related, this more than likely means that a failure is going to result in a system that is going to require in-place servicing. Considering the relatively small size of the system, even a modest size farm would likely require many bots, so failures are bound to be a frequent occurrence. Ideally, it should be possible for the most common failures to be repairable in place quickly with minimal specialized training.
Any failure other than the power system, control system or drive train would be addressed by commanding the system to return to the shed for repair, while sending out a “spare” unit to continue the job if one is available.
A design with readily-accessible, interchangeable, standardized modules would make drive failures much easier to repair in the field using on-hand spares. The standard kit for a maintainer should include one complete set of drive components, with a quick swap-out in the filed and return of the failed module for “depot” service (which might just be the barn or storage shed to either be repaired locally or returned for service). Bonus points if the only simple tools are required to do the job.
Now, having said that, the drive system modules are definite candidates for DFMA, but even then the design for an agricultural application should favor user serviceability and modularity over lowest cost of manufacture. I haven’t thought it through enough, but it seems to me that at the Acorn’s size (and up to maybe small-scale truck size) that a single steering system to control pairs of wheels should be sufficient? Individual wheel drives still seem to make sense, though.
Various NASA designed Mars rovers, designed specifically to handle uneven terrain, with individual drive and steering motors per wheel would beg to differ.
Heh heh, stirred it up did I? Well, for the given earthly farming task on sane earthly farming land (see the demo) it only requires a single steering motor driving minimally 2 wheels locked by linkages. Messy farming terrain? Well there are still 4 independent wheel drives. So, all that added cost, including pollution to create the 3 extra motors, the extra power capacity required, etc. Also, as pointed out, more (expensive) parts to maintain. The efficiency coefficient I quoted was just nonsense out of thin air, but design does as noted. If you still think I’m wrong then I’ll point you to any farm tractor you can find.
This is a perfect mate for “J5” (from BlankMan)
Rusty, very interesting and valid points, with one question – why do you believe there would be extra power capacity required, unless you are thinking about the very minimal power requirements of the additional motor controllers?
Well, my bad, digging deeply enough (really, open source, but you have sign up for an online cad account) they do only use one steering motor, and _without_ any linkages to other wheels (that’s likely a poor choice).
As for the question about extra power requirements, more motors equals more power not only due to the extra electrical loads, but also due to the extra weight. Remember, gravity sucks & frickin’s a drag. 9)
There is 1 thing it doesn’t need to be beefy to do. Follow the trace wire that irrigation pivots use and find a break. Right now that’s mostly a guy walking in a field with a tracer taking periodic readings. Sometimes a day or two if it’s really sneaky. This could go along and mark any breaks or any locations where the insulation has been breaking down but hasn’t shorted to ground yet.
Wow.
Open source project for standardizing & enhancing guidance & control of an AG robot.
Hope the market finds & rewards you.
Million dollar idea.
Unlimited potential.
The fact that it’s light is a very good thing, it has a smaller footprint than a human and doesnt damage the soil and it’s inhabitants.
A line of 3 or 4 wire cutters in front could be great, with some kind of light harrow or fork that raises when an 80 cm height is obtained with the cuttings… But I’m having pleasure using my scythe, it makes no annoying noise and can be very selective.
Thanks for making this open source! Making it autonomous in steering and navigation for its basic platform will make it easier to add on. For example, targeted foliar spraying or fertilizer applications would easy to add. The weeding of small weeds can also be done early on and there would be no need for chemical pesticides. Harvesting produce could come later as the system becomes more sophisticated. I would add powerful batteries so that it can also recharge itself. Adding cameras while recording in cloud server and issuing alerts if intruders approach it would help apprehend thieves.
Here’s a fun hypothetical question. Might lead to some product or project ideas.
Lets imagine running a small fleet of these systems on a farm. Other than the capabilities of the systems themselves and their basic operation, what other issues would you need to keep in mind and actively manage? Some things that come to my mind are:
– Planning. How do you coordinate multiple bots with different attachments to do meaningful work over a large area?
– Safety (don’t want the machines to run into people, animals, or other things)
– Coordination (don’t want the machines to run into each other)
– Security. As other comments have mentioned, this looks like a juicy target for theft or vandalization. On the other hand, the could also be used to keep an eye on the property and each other.
– Maintenance. What kinds of service equipment, spares, and maintainer training would be needed? If a system encounters a problem, could you retask other units to fill in while its being fixed? Automatically send out a ready-spare system? How much self-test capability should the machine have? How does it report a problem?
– Weather. Lightweight machines with solar panel roofs. This sounds like something that I would not want out in high winds. Would you incorporate weather monitoring in the control planner or just trust that an operator was overseeing the area and could respond manually? Do you recall all units to the shed when bad weather is likely or would you build small shelters in the field for them to retreat to if something comes up?
– Communications: How much information do you have to send to control the flock? Would you want to install comms gear in a few locations on the farm, or maybe use a mesh networking scheme between the units? If mesh, what do you do if a critical node drops out, leaving some systems out of reach? Do you trust a single communications system or include a lower-rate but longer range backup system? What about detecting interference or malicious tampering with the comms system?
Sounds like an interest mission management tool might be helpful.
And finally, a shout out to the start of the 1984 movie Runaway with Tom Selleck. It starts with a malfunctioning AG robot.
There is a commercial farm robot like this, Californians, you are (onece) too slow: https://farmdroid.dk/en/case-stories-3/
At 800Kg and 6-8 row coverage, the farmdroid is not designed for the same scale of operation as the Acorn and similar small-scale systems. In AG, the field is ripe (…sorry…) for innovation at many levels of operation and price points. From the pictures on the site, the farmdroid is probably meant to work a the same level as a ?medium? tractor, with a cost of ownership to match (its purchase cost can be a lot more than an equivalent tractor, as long as operating costs are lower to offset the investment).
On the other hand, a single, larger unit may make sense even at smaller scales. Its big enough to cover several rows at once, and looks large enough to handle winds and weather (and looks too big to easily steal). Smaller systems may offer an advantage in flexibility and scalability, though.
Question for the developer (or anyone else with experience).
How do you make a robot rugged/weatherized? On a farm, the robot could experience dirt, mud, rain, sprinklers, etc.
I only have experience with educational setting robots, never tried to make something that intentionally got dirty, and wouldn’t know where to start. I mean, tractors (and cars) have to contend with this, but from a development stand point what do you design for/procure from?
There are a lot of NEMA outdoor electrical waterproofing components and enclosures available now. There are tradeoffs with processor/electronics cooling requirements, of course.
https://www.polycase.com/waterproof-enclosures
Great 👍
The agricultural industry has not always been enthusiastic about the implementation of advances in technology and technology, and this traditional way of thinking is beginning to change. I am very impressed with the mission of the company Slopehelper “No more hands in agriculture”. Agricultural robots https://slopehelper.com/how-it-works are the future of industry