Acorn, our open source precision farming rover
community.twistedfields.com
community.twistedfields.com
https://community.twistedfields.com/t/march-2022-update-simu...
A common question is “okay but what does it do?”
The answer is that right now it doesn’t do any farming. To do farming we need to develop tools. Tools are highly specific to the farming process used, and our machine is unlike a tractor or a human hand as it has less power than a tractor, more power than a human, and doesn’t get bored or require any marginal input costs like human labor does. This means existing tools aren’t exactly right and it’s going to take a bit of work to get that sorted.
Because tools are such a complex part of the system, our plan is to finalize and release our autonomous rover without any tools to early adopters, and then work with the community to experiment and design tools together. Remember that this is an open source research project and we encourage and will assist anyone in going in to production on their own machines, so early adopters designing tools will ensure they have a complete system they can manufacture that works for their farming context.
The end goal is a system which can help in every step of the farming process in a regenerative organic farming system. This includes clearing the rows of any cover crop, planting seeds and starts, thinning seedlings and pulling weeds, dealing with pests, harvesting, and resetting for a new rotation.
It’s a lot of work as I’m doing most of the design, engineering, and fabrication work myself as well as all video production and blog work, but as I’ve mentioned I’m finalizing two new videos and one of the new videos is a new easier to produce format. I got a little too in to fancy video production with four locations and multiple angles on one of the videos and realized I need to dial it back and produce simpler more frequent videos.
You’re welcome to follow me on mastodon where I post photos of some of the latest work, as well as personal stuff. Here’s a picture of the brand new vehicle we have developed that is our production intent system, as well as new electronics which have just finished production and are in transit to us now. Thanks so much for taking a look!
Well yes, the main thing I am expecting when I see a video about a robot/rover - is simply to see it in action. In the 9 minutes video it is shown to be moving only some seconds, which I totally missed upon skimming through it, which left me disappointed.
So if it really can do GPS pathing, I want to see that, so why not show it? Probably best in a speedrun, where it goes up and down the field.
All the technical details are interesting, but probably are better shown after it is shown what it can do. Basically all the robot movies are following that pattern and it makes kind of sense. Because you likely want to attract people who want to use it and not mainly people who want to tinker with it for the sake of tinkering.
"Tools are highly specific to the farming process used"
Seeding and weeding (in early stage) are quite general problems (mostly) independent from the plant and especially autonomous weeding (to not have to use pesticides) could be really useful. You are probably aware of https://farm.bot? They are doing it stationary, but maybe the same tools with modifications could be applied to your rovers?
The new video that has more complex production is an 18 minute narrative video sharing the full details of our intent as well as a story which has been important for me to share for a long time. To keep it engaging I decided to shoot in a few different locations at the farm and change camera angles, but I learned that it’s just too much to attempt to do that. So future videos will be me standing in one spot reading an update while I play b-roll of whatever we’re discussing on screen.
Weeding tools are going to be more complex and rely on computer vision, which requires building a data set and a lot of complex mechanics, so we want to release the system to early adopters before that. Plenty of people are interested in this, and anyone who wants to wait till we have tools is welcome to do so. It’s just going to take a while because we’re not venture funded, we’re a little open source research project.
Here’s a video of autonomous operation. Also in my first link in the comment you’re replying to there’s a video showing complex multi-path following capability. Cheers!
"So future videos will be me standing in one spot reading an update while I play b-roll of whatever we’re discussing on screen."
That sounds solid enough. I am also in the process of working up the presentation of a software open source project - and I totally get lost in fancy details, while forgetting the essentials.
Anyway, keep it up, looks good!
I would love a tool like this, even if it required setting up some sort of track, which could run a standard seeder on a precise path and then habitually run a standard cultivating hoe within a quarterinch of the rows it seeded without me thinking about it. I could never rig up something suitable on a BCS toolbar, its bed clearance wasnt enough. Being able to not interact with a crop beyond irrigation maintenace, filling seed hoppers, and harvesting/bed flipping would be amazing.
For example, hemp is a crop that is fairly easy to plant and fairly easy to harvest (albeit much more labor-intensive to trim and make into a final product), but it's relatively hard and time-intensive to look for mold or aphids and mites, especially as its get closer to harvest. The nice thing about this goal is the tools needed are just cameras and software on top of your existing platform. (These are my opinions as someone who has farmed hemp commercially, albeit only for 2 years on 6 acres).
It'll presumably have to carry and process the harvested materials too.
1. Weeds. My uncle's old cultivator sits rusting in his machine shed because he just sprays herbicides these days and running a mechanical cultivator is slow, weather-dependent, and hard to scale to hundreds/thousands of acres across several fields with different locations. Mounting a mechanical cultivator between the wheels on either side or devising some other system of weeding would greatly reduce the amount of spraying, and of course who cares if Acorn is slow? Set it and forget it, have it go faster when no weeds are detected, focus on where the weeds and wet patches are. Farmers don't like spraying herbicides, but they do it because it's legal and it's the only way to scale to current acreage.
2. Fertilizer. So much fertilizer is wasted and turns to runoff. Applying fertilizer precisely would be a huge boon to the Mississippi and the Gulf.
3. Irrigation. Some farmers in Wisconsin irrigate, but most in Iowa don't. Some parts of a field are wet, some are dry, but the air is usually humid. Using condensate to irrigate crop roots directly in dry areas of a field could prevent a ton of loss for farmers.
The solar panel and supercapacitor architecture is interesting, but it'd be great to see what could be done with some batteries (an array of ebike batteries?) that could be charged by the panels. With the assembly, allowing adjustment in the width would be useful to accommodate different row sizes, height clearance to accommodate different crops, and then mountpoints between the wheels on either side for different equipment/attachments. I'd think some triangle braces from the wheels to the side centers may be necessary.
It'd be great to see how well Acorn could deal with different grades and mud. There's increasing evidence that heavy tractors have a long-term detriment to fields due to soil compaction, so having lightweight robots would be very useful.
>We are often asked: what does our robot do? At this time Acorn is still under development. In the future it will be equipped with a smart vision system and robotic tools that will be able to plant seeds, destroy weeds, monitor plant health, and much more.
But so far all it looks to be is a solar powered buggy. Would love to revisit this when they make it actually do something.
Claiming it's a "precision robot" before claiming it does anything is...a choice.
But it is supposed to be only driving when the sun is shining, so maybe we have to wait till spring/summer?
I mean seriously? It surely is more efficient, to not waste solar energy by storing the energy in capacitors and not in batteries, but that means, it will stop whatever it is doing when a cloud comes and it will stay out and not make it to the shed again, when the sun does not come back.
I would rather have the solar panels somewhere else, or those on top of the rover just as a bonus and use batteries and focus on it to do useful things first.
edit: ok it is driving in the video at 8:25 for a few seconds
The solar is an experiment to see if we can build a system that reduces waste. Farms already rely on sun to operate and we want to see if we can build a farming system that does the same. This reduces waste as our super caps last a million charge cycles while lithium batteries only last a few thousand.
Some other guy claimed his robot was "full self driving" and was quite successful with that choice ...
https://news.ycombinator.com/item?id=34167102
I will say that while it doesn’t do farming yet, that is the point of the system. And currently it autonomously controls 8 motors to follow precision GPS paths based on data from a central server and database system with a web page for control. It’s definitely a robot with some degree of precision!
1) Precision weeding. The best option I've seen so far is using an IR laser and machine-learning based weed-recognition software to weed. Here's one example:
https://www.freethink.com/robots-ai/farming-robot
> "The Autonomous Weeder can eliminate more than 100,000 weeds per hour and weed 15 to 20 acres of crops in one day — for comparison, Myers said a laborer can weed about one acre of his onions per day."
The other has to do with precise fertilizer application. Right now, it's cheaper labor-wise to just dump fertilizer on the entire field, where a good fraction of it is lost due to things like runoff or denitrification. A robot could instead deliver aliquots of fertilizer right to the base of each plant, and could possibly even monitor plants individually for deficiencies. This could vastly reduce fertilizer demand in the optimal situation. This is also under current development:
https://www.frontiersin.org/articles/10.3389/frobt.2022.8084...
Nice to see people developing DIY versions, too! (Though I imagine CO2 lasers are a bit rough on the budget).
Residue free, low effort weed control. The future is here
Taylor, I was just wondering-- Is it feasible to use lasers for killing weeds while retaining relatively light weight robot? (Wasn't sure if this would require prohibitively heavy batteries)
Here are some other interesting farm robots--
- Six farm robot models-- created by folks at the University of Sydney, Australia
https://confluence.acfr.usyd.edu.au/display/agpub/our+robots
- Carbon Robotics' "Laser Weeder"
When you're manually moving the tractor, do the birds ever get stuck/pinched under the edges? Seeing videos of them being moved, it seems like it might be easy to have happen if the chicken got lazy.
* QuT AgBot II is still active (2016) [1]
* RIPPA (2018) is working commercially while being advanced [2]
* In cattle management Swagbot (2018) shows promise for future roles [3]
[1] https://www.youtube.com/watch?v=XXs21541-Vg
Why lug those huge heavy panels everywhere?
A modular design, where panels are installed on the farm, and the heart of the machine is small, compact and has different chassis to suit different farming requirements, is a better way to go forward.
The solar panels charge a swappable battery and the farmer can switch batteries every two days or so.
Currently, the width of the bot is an issue. Its too big. You want small bots that can go between plants close together and do their magic.
I'm sure there's a concern that the panels are undersized, especially when it's working in mud and trying to dig holes on a cloudy day.
What is the ideal row spacing in your opinion? Row spacing is a factor of the size of the equipment (tractor and various devices they pull), and a size of the plants.
Row spacing requirements vary greatly between different types of farms.
For example, the current configuration is not suitable for corn or maize once they grow fully.
My ideal case is to have a box that is pluggable to different chassis, that can be used at different stages of farming and at different types of farms.
Google says 100W panel is ~14 pounds, so in this case 800W / 112 lbs.
Assuming you're getting an average of 800W over 6 hours: 4.8kwh per day, or 9.6 kwh for a couple of days.
Using Tesla packs, 100kwh/1,300 lbs implies ~ 125 lbs. in batteries.
Not an obvious win just on weight.
It makes sense to me that a single person can do farming by hand with about the same power available as our robot. There’s no heavy tillage in regenerative organic so the power requirements are much lower.
Also the nice thing about the solar power is that the robot is just always ready to go. No charging, no battery swapping. No need to source electricity from anywhere. All you need is a modest amount of sunlight and you’re good to go. Honestly when we started the project I wanted to do batteries, and Daniel insisted we do solar and supercaps. After building it and using it I’ve fallen in love with the system. It’s so nice! The trick is to co-develop a farming system and the robot to get the farming done within this power envelope. We believe it’s likely possible. We can always slap a battery on it if needed. And I do enough software development at night I do have a battery for testing. But I’d prefer that be used for special cases rather than every day use.
There's a reason that tractors are heavy.
https://upstreamaginsights.substack.com/
https://www.nature.com/natfood/
https://www.frontiersin.org/journals/sustainable-food-system...
Future of Agriculture Podcast
Red to Green Foodtech Podcast
Would be more productive to force big manufactures to open-sourcing their stuff.
Given that current manufacturers' agtech tends to be incredibly locked down devices, just about anything starting as open source seems to be a much better use of time than trying to force John Deere, et al to open source their devices.
There are interesting developments around using precision targeted spraying at individual weeds that this robot would be good at. The chemical volume and hence weight would be low, so you don’t need a big diesel tractor!
Similarly, a small robot like this could make daily passes over a field, killing weeds when they’re still tiny. Again, this reduces the power and weight requirements.
taylor needs immediate hands on and real world farm work experience to try and not only solve a problem but figure out what farms need. most of their efforts feel out of touch...in my personal opinion theyve consumed too much of the marketing of farming. cheap open source implementations are a big plus, but theres a reason we dont use bicycle tires on tractors or equipment. light weight can also be a huge problem after a rainy season or during a storm. oklahoma farms get 50mph gusts and texas farms need to operate +35c. can we survive a lightning strike? if we get lost can we alert other farm traffic were around? can we get accidentally sprayed by a chemical or forgotten somewhere? can we detect a person?
taylors robot needs to run on numerous fuels, including diesel or kerosene, to serve small farmers globally. it needs to be a lot more hands on, easy to transport and quick to get visual information from in english and especially spanish.
Edit: so ive given it a little more thought and in picking season might be a good candidate!! can we use Taylors robot to detect fatigue or heat stroke? lightning strike detector? live dopplar? respond to accidents, communicate first aid? deliver food and water, or confirm a crop for picking? in short i think taylors offerings a lot more on the human side of farming, like a digital foreman :)
I should point out that we’re intending to target regenerative organic farming systems, which are often tended by humans with hand tools or small 15 horsepower two wheeled tractors (like a BCS). At that scale, our one horsepower solar powered system can likely be made to work. But we don’t have to solve every problem ourselves. We’re trying to get this in other peoples hands and the design will evolve as we get feedback from real world farmers and experimenters.
Edit: This is the type of farming process we’re targeting. Hopefully that makes it clearer why we believe in our approach: https://youtu.be/1BH0NkN6zHs
dont discount the wind though! many farms see a lot. having onboard diagnostics like a printer does might be nice too...quick things like spot weld repairs?
Yeah we may need to add gaps to the solar panels so there’s less surface for wind to grab, and we’ll have to make sure they’re secured nice and tight! I hope we get some customers in windy areas to test and give us feedback too. It’s all about community involvement here.
Diagnostics would be great and we do have some of that rolled in to the software now. I’m sure there’s room for more. We are MIG welding the new frames and they should be quick enough to do in the field if needed!
Personally I am not a fan of the way the project is presented, though.
not a lot can be done about it with the feedback you provided though
Some plants (even native/noninvasives. looking at you Lupine…) cause serious harm directly to livestock, or reduce the carry ability of the pastures (mustards, St. John’s wort) to get livestock to late fall without eating the hay you cut in the summer that might have a cluster of hoary alyssum which will make your horses sick. Some, you may just have to make your peace with (Canadian thistle).
Plants are scary.
Monomaniacal fabricators trying to cover the entire surface of the Earth with copies of themselves.
... which means the Acorn is powered by an Acorn (RISC Machine) CPU.
Hacker News: Come For The Hippie-Bots, Stay For The Jokes Based On Obsolete Acronyms