FarmBot - Open-Source CNC Farming

We’re gearing up for our big launch in July and wanted to send out a little tease of what’s to come. Check out the FarmBot Teaser Trailer:

Check out this FarmBot Q&A with Edible Manhattan, Food Loves Tech, and AgFunder! The original article can be found on AgFunder here, and it is copied below.

Editor’s Note: Suzanne Zuppello is a journalist with food magazine Edible Manhattan, and has been writing about food tech in advance of the Food Loves Tech conference next month.

Rory Aronson wants you to steal his idea. Seriously. In founding FarmBot, he decided none of his ideas would be proprietary. Everything he and his team design and all of the information they gather is yours for the taking. It’s that strategy that keeps the FarmBot team constantly innovating and reacting to new ideas and feedback to make the best, automated farmer for your backyard.

Want to learn more about FarmBot? They’ll be in New York City June 10-12 for Food Loves Tech.

Suzanne Zuppello: FarmBot sounds like the title of a George Lucas film. Is it as sci-fi as it sounds?

Rory Aronson: Sort of. After graduating with a degree in mechanical engineering, I decided I wanted to reinvent the way food is grown in order to adapt to the growing use of technology in people’s lives. When I thought about rebuilding the agricultural process, a robot is what I pictured. It is quite literally a robotic device that someone can control through their smartphone or laptop. It’s simple enough to use in your home but sophisticated enough to adapt to a larger scale.

SZ: You say anyone can use FarmBot—but I imagine some knowledge or skill is requiring before diving into building a backyard farm.

RA: Not necessarily. A person decides what they want to grow, and FarmBot will do the rest. It can plant seeds and systematically water them based on how old the plant is, how it’s growing and what the local weather is. Much like having a 3D printer in your home and deciding what you want to make, you can decide what you want to grow to your own specs and FarmBot will do it. It transitions the control of the system one may have to the consumer. For instance, if you want to grow spinach in Manhattan or kale in the tropics, FarmBot can gather this data in order to grow the plant.

SZ: But where is FarmBot gathering it’s information from?

RA: We’ve created a non-profit database called OpenFarm that’s linked to FarmBot. Through this, people can upload growing guides for trees, herbs, flowers, mushrooms — any agricultural product. It’s like a recipe book for farmers. Anyone with or without a FarmBot can use this data and learn how to garden. This data is also available to any agricultural technology through an open API.

SZ: It’s an interesting business model to give information like that away for free.

RA: My drive in establishing this company is not profit. Everyone eats, so everyone should have the ability to grow their own food, with a little help. By making our growing guides as well as the specs to build a FarmBot totally open sourced, we’re acknowledging that we think we’re doing it the best while still being open to improvements entirely. If someone can do it better than me, I welcome that because that will push our mission and technology forward.

SZ: So I take it there’s no patent pending on FarmBot?

RA: Not at all. I’m not worried about the lack of a patent. Proprietary products can allow for competitive advantage, but it’s my belief that it’s a cheater’s way out. My competitive advantage is that I’m the first to market and building a brand of trust that puts consumers first. I believe open sourced information is the way of the future. More companies are opening up about their practices and technology, in order to put the consumer first. I think people are going to respect the way we’re doing this rather than a competitor who takes everything from FarmBot and then makes it proprietary and doesn’t make it open.

SZ: That’s an incredibly honest and valuable approach. Is the machine ready to ship?

RA: Not yet. We’re still in R&D mode. There’s actually a prototype moving through my yard right now. I have cameras pointed on it and the plants to see where we can improve. Ultimately, we’ll create a timelapse of a three month period showing everything from planting to harvest with a FarmBot. Once this video is ready, we’ll prepare a crowdfunding campaign to invite people to join the team as early adopters and implore them to hack our system in order to improve it. When it does ship, the user will need to assemble it.

SZ: Well, I hope it’s easier to put together than my Ikea bookshelf!

RA: We think it is. Every piece is cut to size, and every wire is there with clear direction. We want to empower people to grow their own food, so confusing them with advanced mechanics is not the way to get there. With that said, we also see FarmBot being used to grow plants beyond consumption. Because the process is so controlled, you can use it to grow plants for lab testing and remove the element of human error.

SZ: Do you see FarmBot as a means to end food scarcity?

RA: While that would be fantastic, it’s not our aim. We are focused on anyone who wants to grow food in a way that requires less energy, less transportation, and hopefully less water and time. Having FarmBot is essentially giving everyone an extra set of hands — and if you have hands, you can do anything.

What do you think? Would you use Farmbot to grow your own produce? Get in touch: Louisa@AgFunder.com

FarmBot is 100% open-source, meaning that anyone with an internet connection can look at and download all of oursoftware, hardware plans, and documentation so that they can build and operate their own farming machine. But in reality, will people actually do this? Are the open resources we’ve created good enough for someone to build their own device? Is FarmBot worth replicating? Is being open-source a worthwhile cause? Up until just recently, we had a hunch that the answer to all these questions was a resounding yes, but we had no concrete evidence in the case of FarmBot. What we did have going for us was the fact that replication already happens with other open-source hardware and software projects. But still, it had yet to be proven with FarmBot.

Today we can definitively say that the answer to all of those questions is indeed yes: people will replicate FarmBot, the open resources we’re developing are good enough, and being open-source is without a doubt a worthwhile cause. That’s because over the last few months we’ve been very excited to see and hear of four more FarmBots being built by people and teams outside of our core development group. These people independently came across our idea, our documentation, and our software; and then decided it was a great idea to build a FarmBot on their own, and went for it. Two of these other FarmBots are in the United States, and the other two are in Japan. Two are being built by individuals, while two are being built by companies small and large. Check it out here: https://farmbot.io/2016/04/30/is-farmbot-worth-replicating/

This is a huge milestone for our team and The FarmBot Project at large, and we can’t wait to see the FarmBot community continue to grow. Want to build a FarmBot? Get started on our documentation hub. Want to buy a FarmBot? We’ll be accepting pre-orders in July. Sign up for email newsletter and you’ll be the first to know when they’re available!

About 8 months ago we decided to completely refactor the FarmBot web app. Today I’m please to announce that we’ve finally finished this system wide re-write and the results are a faster, more stable, more well-written/tested/documented and ready for the future version of the software that powers FarmBot. Here is a summary of the changes we’ve made:

• Switch from a traditional multi-page web application to an SPA (single page application). This allows the web app to load “once” during a session and then not require any page reloads as the user navigates through the app. This makes the application faster to use because everything is loaded and there is not as much network lag or abrupt page refreshing.
• Switch from Angular to React. We weren’t too happy with Angular and its performance with our specific application needs. We decided to switch to React because of its already large and quickly growing user base, as well as its relatively easy learning curve.
• Switch from Foundation to Bootstrap. Foundation was proving to get in our way which required us to override it often and it wasn’t playing nicely with Angular. We decided on Bootstrap because of its ubiquity in the developer world and larger community of support. As it turns out, we’re not using very many of the components that Bootstrap gives us, so we may end up ditching it too for a purely custom solution or for something that provides things we want but could not feasibly build ourselves.
• Switch from Meshblu to MQTT. While Meshblu communication got us up and running quickly in the early days, it began to cause trouble later on. Features we were using (such as logging) got removed, upstream changes for websocket support silently broke our systems, and we were at the mercy of a complicated corporate controlled codebase. Because MQTT is a widely adopted standard with many more simplified brokers out there, we decided to make the switch. Its been much easier to setup, debug, and maintain our systems now.
• Modularization of the frontend, backend, and farmbot.js. Before we had one monolithic codebase called farmbot-web-app. It was difficult to maintain and programmers (especially newbies) had to get the entire thing working to be able to make small changes. For example, to change some css and see the changes in the development environment, one would need to get the entire database working too. Now we have three smaller repositories that handle specific functions. farmbot-web-app is just a backend system and API (the business logic). farmbot-web-frontend is the entire frontend which is compiled into a single javascript file via react magic. And farmbot-js which is a library enabling the frontend to communicate via MQTT with the FarmBot device.

Phew! Lots of changes huh? That’s why it took about 8 months to do this huge refactor. But now, we have a better software system that will be able to grow with us into the future. Want to see the new app for yourself? Check it out atmy.farmbot.io or see the screenshots below. Note that a lot of features still don’t work, but the foundation is there :). Over the next few months we’re going to be laser focused on getting sequence building, regimen building, and the farm designer working so that we can demo those features at the time we launch later this year!

We just wrapped up the documentation for FarmBot Genesis V0.9! In this version we added interactive virtual tours each sub-assembly, more detailed assembly instructions and troubleshooting tips, ideas for FarmBot mods and hacks, docs for new tools, and tech specs for every component in FarmBot. These are our most detailed docs yet, and we can’t wait to see you use them to build your own FarmBot. If you have any questions or issues, drop us a line in our support forum. Best of luck!

People often want to know how much electricity FarmBot uses. Sometimes they’re wondering if FarmBot can be run on solar power (which it can be). Though more often they are actually searching for an answer to this (slightly more complicated) question: “What is the carbon footprint of owning and operating a FarmBot to produce vegetables compared to buying an equal amount of vegetables from the store?” In this post we’ll look in depth at this question from a theoretical perspective.

The two types of emissions

First let’s talk about carbon footprint calculations. In general, there are two types of carbon emissions associated with a product:

1. The emissions generated from producing the product. These emissions come from mining, processing raw materials, manufacturing the components/product, and shipping it all to the consumer. These emissions are usually directly related to the embodied energy of the product, though depending on the source of the energy used in production, the total emissions can vary widely. For example: the aluminum used in FarmBot might be from Factory A which uses solar energy (zero emissions), or from Factory B which uses energy from coal. Because we don’t know every detail of the FarmBot supply chain, we don’t know exactly what sources of energy are used to produce a FarmBot. Instead, we will be using published averages for the calculation of these emissions.
2. The emissions generated from using the product. FarmBot uses electricity to operate. The source of that electricity will determine how much CO2 is emitted to operate FarmBot. Because we don’t know if the consumer will be using coal based energy, that from natural gas, wind, solar, or some combination thereof, we will again be using published averages for electricity usage emissions.

Sometimes a third emission is included which accounts for emissions generated after a product’s lifespan. For example, a building might cause more emissions if it needs to be demolished and taken to a landfill when it is time to be replaced. For this post, we are not going to be considering these post-lifespan emissions as FarmBot is a physically small device (compared to a building) and any emissions generated from recycling materials would be attributed to the new product, not FarmBot.

Embodied Energy and the Emissions from Producing FarmBot

In the table below we list the most prevalent materials used in FarmBot, their cumulative weight, and the expected kg of CO2 emitted due to their production. Keep in mind that these are only estimations meant to provide us with a ballpark idea of the emissions generated to produce a FarmBot.

It is important to note that the numbers in the table above are calculating the CO2 emitted for the production of enough raw materials for a FarmBot (raw aluminum billets, raw plastic pellets, etc). Obviously FarmBot is not constructed from raw aluminum or raw plastic – it is constructed from aluminum extrusions and plates, 3D printed plastic parts, screws, wheels, belts, motors, and other non-raw components. In reality, the actual CO2 emitted to produce a FarmBot may be closer to 150kg (330lbs) than 100kg, though we can never truly know without exhaustive research and due diligence.

Emissions from Using FarmBot

In this table, we look at the emissions attributed to using FarmBot. While we could consider the emissions associated with the delivery of municipal water, acquiring seeds, and more, we are going to limit our scope to the most obvious emission source: electricity usage. Because we don’t know how a specific consumer’s electricity is produced (coal vs solar vs whatever) we’re going to use the US average for CO2 emissions per kWh of electricity: 0.554 kg CO2/kWh.

Note: Duty cycle refers to the estimated average percentage of time that the component is being used. For example, the computer and microcontroller are always on (100% duty cycle), while the motors are only being used sometimes (estimated at 5% for each motor). As a point of reference, if you have your TV on for 4 hours per day, that is a 16.6% duty cycle. Also keep in mind that we’re including usage of high power 12V tools utilizing the maximum deliverable power, and also an always-on webcam. At the end of the day, these are very rough estimations and the power usage may change considerably depending on how you program and use your FarmBot.

Using these estimations, operating a FarmBot will emit approximately 60kg of CO2/year, or 130lbs. As points of reference, this is equivalent to burning 6.5 gallons of gasoline in your car, and the energy used each day by FarmBot is close to what a desktop computer would use in about an hour and a half. Fun fact: at the average US electricity cost of $0.15/kWh, operating FarmBot for a year will cost about $16.

FarmBot Veggies vs Store-Bought

Now that we have a ballpark estimation for the CO2 emissions of FarmBot, let’s see how using FarmBot to grow vegetables compares to buying the same quantity from the store. From our analysis of how much food FarmBot can grow, let’s assume we’re fully utilizing our FamBot and growing 80 calories of vegetables per day. This means that over one year, FarmBot will produce 29,200 calories of vegetables at a cost of 60kg of CO2. This means that the carbon intensity of FarmBot grown vegetables is 2.05 g CO2/calorie.

According to ERS/USDA data seen in the figure below, the average carbon intensity of vegetables grown in the US is 2.8 g CO2/calorie.

This means that FarmBot grown vegetables emit approximately 25-30% less CO2 than vegetables you can buy at the store. Keep in mind that this is not taking into account the CO2 emitted for the production of FarmBot itself (150kg), though the CO2 emitted from the production of industrial tractors, transportation trucks, highways, and grocery store buildings are not included in the 2.8 g CO2/calorie number either. Considering the difference in materials used in the US food supply chain vs the materials used in a FarmBot, we can only imagine that taking all that into account with an exhaustive study would only prove FarmBot to be an even more attractive food production system from a CO2 emissions standpoint. So let’s add FarmBot into the plot! Fits in nicely, huh?

Other Things to Note

• This analysis is assuming you use average US electricity. If you run your FarmBot off of solar energy, then the CO2 cost of FarmBot grown veggies goes to near zero g CO2/calorie! The only emissions would come from the usage of the web app server and your laptop/phone (unless your whole home is solar powered and you run the web app locally!)
• Because FarmBot grown veggies are grown in your backyard, there will be no carbon cost associated with getting the veggies from the field to your fork. With store-bought veggies, you might normally drive to the store – an emissions source not accounted for in our data from the ERS/USDA. And that source could be huge. Remember, the amount of emissions FarmBot emits in an entire year is equivalent to burning just 6.5 gallons of gas. If you drove to a grocery store or farmer’s market 2.5 miles from your home once every week in a vehicle that gets 30 mpg to pickup veggies, then you would burn over 8 gallons of gas in a year, effectively doubling the carbon cost of those store-bought veggies.
• Because of these reasons and the conservative numbers we’ve used in our calculations above, we suspect that in reality FarmBot grown veggies will outperform store-bought veggies by an even greater amount than this analysis has suggested, especially if extra steps are taken to make FarmBot less carbon intensive, such as going solar.

Sources Used

• http://www.ecocem.ie/downloads/Inventory_of_Carbon_and_Energy.pdf
• https://www.carbonfund.org/how-we-calculate
• http://shrinkthatfootprint.com/food-carbon-footprint-diet

For a while now we’ve been planning to launch on Kickstarter with two versions of FarmBot Genesis: a “standard” sized device measuring 1.5m wide, 3m long, and 0.75m tall; and an “XL” device measuring 3m wide, 6m long, and 1m tall. In fact, the most recent version of the hardware prototypes (V0.9) have been designed with larger plates and extrusions, and more V-wheels to help strengthen and stiffen the tracks, gantry, cross-slide, and z-axis for the larger sized form factor. The design intention is that one set of hardware components could be used to build either sized device, or any size in between – a scalable FarmBot. The only modifications would be the need for longer cabling, tubing, and belts; and extra extrusions and plates for the larger capacity.

However, even with the latest prototypes, we have decided that we would rather launch with just one sized device – the “standard” sized FarmBot Genesis. Here are our reasons:

• Simplicity – With two devices comes two kits, two sets of packaging, more components to order and have on stock (longer wiring harnesses, etc), two sets of assembly instructions, marketing two slightly different value propositions, and providing customer support for two devices. We’re a small team with limited resources. Adding a second version of our flagship device to our launch campaign and business operations will undoubtedly be more complex and difficult than launching with one device. We’d rather start small and “do it right”, than try to “do it all” and fail to do anything well.
• Speed – A more simple product launch of just one device will allow us to launch sooner and ship devices faster. It is easy to fall into a trap of continuously adding features and making changes in lieu of shipping what you already have developed. Launching sooner will allow us to gain feedback quicker, build the FarmBot community and expand our team faster, and iterate more rapidly.
• Engineering Reliability – While we think the design intention for a scalable device is good, it is only feasible to a certain extent. Right now, the FarmBot Genesis hardware scaled to the XL size does not appear to be reliable – there is still too much flex within the system, and other challenges begin to arise (such as thermal expansion) when the hardware is pushed to these scale limits. In order to solve these scale challenges, there need to be more serious modifications made to the design likely including a switch to larger motors, an A-frame style Gantry structure, and a different material choice for the tracks. Such modifications however would then make the smaller devices significantly overbuilt, pricier, and less functional. Ultimately, we think that a FarmBot the size of Genesis XL will need to be a completely separate design that lives up to its name, rather than trying to squeeze too much scale out of the slimmer standard design.

Thank you to Hackaday for inviting me to speak at the Hackaday SuperConf!!