Growbotics AI Vision
I started Growbotics AI project at the beginning of 2018 and wrote the initial blog post to give friends and other people the idea what this is all about. The plan was to write a second blog post about the journey and lessons learned while building a miniature greenhouse. The concept of such greenhouse is rather simple: create one cubic meter sized box, have LED lights in the top, a water tank in the bottom, stuff the cube with many sensors and allow AI (Reinforcement Learning or RL) to control all the knobs. As it turns out water leak is a harder problem to solve than learn how to use RL.
However, work is still in progress, and there’s real hope of getting on top of water leaks, so the lessons-learned post still has its change of seeing the daylight.
In the meantime feel free to check the progress on Instagram.
As time passed the vision behind Growbotics AI project became a bit more clear and mature. I also met a few people who seemed excited about this project and wanted to help one way or the other. I too would love more people to join and help lift this project off the ground, so perhaps a good idea would be to share the vision and see who else is interested.
So let me start with these seven sections, and you’ll have a basic understanding. Moreover, if you’re interested, then a good idea would be to scroll down to one of the sections below and read the detail. The parts are more or less separate, and there’s a summary at the end.
These are the seven main principles:
- Promote crop diversity, urban gardening and growing food locally;
- Recycle plastic, paper, and food waste;
- Make AI-powered robotics more accessible to small and medium-sized farmers and urban gardeners;
- Create a community of people such as data scientists, farmers, makers, designers, software developers, end consumers and others;
- Promote radical transparency, use and create Open Source software and hardware designs, promote the use of Open Data;
- Support using renewable energy by making wind and solar power more affordable;
- Create an innovative business model that would allow people from all over the world to contribute their time and ideas to this project and get rewarded;
Have you caught yourself thinking that perhaps we live in the times of tasteless tomatoes, green skin bananas and picture-perfect apples? If the fruit or vegetable is too exotic for our climate, maybe it will be flown to us by a large cargo plane. To make fruit and vegetable picture-perfect, perhaps the plants will be sprayed with pesticides and herbicides, and fruits picked well before they are ripe. Even if no one used pesticides and herbicides, we can’t know that. From an economic point of view, it’s much simpler and easier to grow one type of lettuce and supply to all the shops in a large city, but that is not so good for the environment.
We can do better than that, we can grow more or less whatever we want, whenever we want and only the quantities we need.
Hydroponics (or a more modern version of it — Aeroponics) isn’t a new technology; however, the cost LED light is rapidly going down. Combine it with the solar energy (or wind if you happen to live in London), and you have an exciting and efficient way of growing plants locally. All year round.
When you can grow many different crops, people’s diet would naturally become more varied, and people probably would be more curious about what they eat. Some people already are doing a great job and are promoting seeds from all over the world. A great example is Dan Jason with Salt Spring Seeds in Canada.
Talking about seeds, we also would want to take the scientific approach. There’s a lot of scepticism around the world about genetically modified plants, and I was initially sceptical about it too. I guess it all depends on what we have to choose between and how informed you are. If the choice is to eat a fruit or vegetable that growers sprayed with toxic chemicals, perhaps genetically modified crop that is engineered to be resistant to certain diseases will do less harm to us. A natural, organic fruit is probably still the best option altogether.
There’s excellent TED Talks presentation by Pamela Ronald about genetically modified crops.
There’s also an exciting development by the Bento Lab team. Bento Lab is the first portable all-in-one DNA laboratory, suitable for applications from professional research to education and it should make it easier analyse seeds and understand what plants we grow and eat.
So this project is all about the crop variety and growing different, more varied crops locally.
We’ve all seen horror pictures of plastic waste in our oceans and on the land. It’s terrible for the planet, it’s awful for the animals, it’s horrible for us, and it looks messy. There’s nothing wrong with plastic itself; the problem is how we treat it. We make much plastic packaging to be used once and then we dump it. Yes, some plastic gets recycled, but certainly, not all and some plastic waste even goes from Europe to Asia to be processed.
I want to mention a fabulous project — Precious Plastic, and I find it very inspirational. Precious Plastic was started in 2013 by Dave Hakkens, and it is a global community of hundreds of people working towards a solution to plastic pollution. I strongly support this project and already bought a plastic shredder and HDPE 3D printing filament made from the recycled plastic in Indonesia.
When it comes to 3D printing, HDPE plastic turns out to be quite challenging to print, but it’s much easier to use PET plastic (factories mainly make plastic bottles from PET plastic). I’ve done some experiments and successfully created a growing pot for the hydroponic greenhouse using 3D filament from Refil. Refil made their recycled PET filament from old PET bottles, and they are proud to say it’s up to 90% recycled! So yes, we can turn plastic waste (or at least a lot of it) into new things. We can 3D print robotic arms, mini greenhouses, small wind turbines and many other things.
Extruding 3D filament where the plastic waste occurs becomes essential — no need to transport the plastic to the recycling factory miles away. Perhaps to maintain the characteristics of the plastic so that it is suitable for 3D printing, some of the virgin plastic needs to be mixed in. A project worth mentioning when it comes to the 3D filament is Filastruder. There is the extruder itself which I already started to test, and there’s a Filawinder, which allows winding the plastic wire onto the spool.
Of course, plastic isn’t the only thing that we waste. We waste food too. Here’s a picture I took at the bus stop in Camden that speaks for itself.
In the search on how to make nutrients for the miniature aeroponics greenhouse, I looked at multiple options. The most attractive option, however, turns to be the composting worms. Some experiments are still in progress, but there’s a high level of confidence that this is going to work. So the miniature greenhouse will be fueled by the nutrients produced by worms from the wasted food. It will be a worm tea mixed with water and sprayed through the nozzles on the plant roots directly so they can be absorbed quickly.
There are some open source hardware designs that could help with scaling-up vermicompost production. It’s available here at Open Source Beehives project, and I’ll say a few words about this project later.
Producing vermicompost still requires some manual supervision, and there are many opportunities for automation. Making a house for worms heated, so that we control the temperature is worth exploring. Also keeping the humidity at the required levels is needed too. Such automation could even lead to the development of the hassle-free kitchen appliance where all you need to do is to put the suitable (for worms) food waste into it, and the machine would create the vermicompost capsules automatically. Having such an appliance could even be profitable: instead of paying for the food waste to be collected, people could get paid for the vermicompost they produce.
We can also recycle the paper. There’s a need to gather a bit more information and do some experiments, but perhaps paper could be shredded, and we could place paper pulp in various 3D printed forms to create packaging. Here’s one very inspirational example shared by the Japanese designer Kazuhiko Hayakawa.
I could also add a few words about used coffee grounds. People love coffee, but after the cup of coffee is prepared the used coffee is often put into the bin. It doesn’t have to be this way, and it turns out that used coffee grounds are an excellent medium for growing mushroom. I can’t report the success on this front, but I tried growing mushrooms, and I will try again, and I’m determined to make it work!
The opportunities to automate agriculture with the use of AI are almost limitless. Instead of daydreaming on what could be done or listing what has been already done by others, I’ll briefly describe the activities that are currently in-flight.
Some of these developments — like The Cube are designed and developed from scratch. Some activities, like the robotic arm or wind turbines, require only 3D printing and assembling. Also, some are just instructions of putting the separate parts together, like the drones or battery packs.
The cube is the primary device that pretty much gave the start to this project. Many people around the world are working on various hydroponics based solutions. Some people or teams focus toward home appliances, some of them towards growing plants at large scale. I look at all these projects like restaurants — some of them are good, but nobody stops you from opening your own, and if done right to some it may be the best. Some designs decisions make these hydroponic appliances different from the ones I already saw.
At the core of the concept, we have a one-meter size cube which is a fully contained miniature greenhouse. The second fundamental concept is that we can stack the cubes onto each other like Lego blocks and this way the structure would share the electricity and water. Such a design allows someone to start with one cube and increase the deployment as needed. This design also allows having something else installed in such structure: for example small size vertical axis wind turbine or a beehive. Another critical design decision is the weight of each cube. It has to be light and weight up to 20kg or at least be lighter than what the heavy-lift drone could lift. Putting one cube onto another can be easy on the first layer, but stacking 5 meters or taller structures is a different story.
So this design isn’t a solution that will allow growing all the possible plants, and the initial focus is towards various herbs. Later, we’ll see where this all leads. Last, but not least thing to mention is solar cells. We designed foldable blinds with laser-cut solar cells and installed them on all four walls. This system can take the required electricity for the LED lights from the sun and use natural daylight when it is available.
There has been a lot of time and energy put into making such a cube, and there will be a dedicated blog post just for it.
The Robotic Arm
The Open Source 3D printable robotic arm — Moveo — has been designed by the company called BCN3D. This Barcelona based company is one of the leading manufacturers of desktop 3D printers worldwide.
So what’s so special about this robotic arm? Well, it’s 3D printable so that people can print it from recycled plastic bottles and it’s less expensive to make one. It’s a general purpose robot that isn’t designed to solve one specific problem. We can use it for a variety of tasks. Here are some examples: sort out the apples based on size for how good they look into different boxes. The bigger better apples could be sold to the end user directly, and the not so good ones could perhaps go into the jam or juice. Moreover, the worms probably don’t care so they would be happy to eat what no one else wants.
Also, what about robotic arm and cucumbers? A few years ago I read, and I got inspired by Makoto Koike helping his parents to sort out cucumbers.
So the goal is to build a Moveo robotic arm, equip it with a camera, use the Robot Operating System, and train it using Reinforcement Learning (AI) to do various intelligent tasks.
The Autonomous Garden Cart
Many people are keeping an eye on driverless car technology. Many companies are competing on who will start offering the first city driverless taxi service as it’s the market which has the money. For such service where there are many participants on the roads safety is essential, in fact, the technology perhaps needs to be more mature than what a human can achieve for it to be rolled out.
The changes in this space are happening quickly. Last year Andrew Ng wrote a blog post and said that self-driving cars are no longer a futuristic AI technology and announced launching self-driving car service — drive.ai. Pony.ai is making rapid progress by testing the self-driving car technology in China and the US where driving culture is very different.
While there’s fierce competition in this space, the technology seems to be mature enough to be deployed in other settings. One example could be in small or medium-size farms where people need to transport various items from one place to another. The inspiration for the hardware design comes from BFF Design Ltd website, we wouldn’t need the handle, and these days it is possible to get the wheels with the built-in motors.
Also, the inspiration for the software comes from Wayve, a Cambridge based company that is focusing on driverless car technology by using only one front mounted camera. I recommend reading the blog posts on the progress they are making.
The plan is also to use Linorobot software. Linorobot is a suite of Open Source Robot Operating System (or ROS) compatible robots. It aims to provide students, developers, and researchers a low-cost platform in creating new exciting applications on top of ROS.
Finally, the garden cart will be only like the base with electric motor wheels and the battery pack. Multiple attachments could be plugged onto it later. Examples of such accessories span from a grass trimmer to the tick picker or garden cargo transporter.
The Areal Photography Drone
Drones turned out to be very useful in agriculture. Some people are using them to spread the seeds, some to monitor the crops. The prices of drones could also vary a lot. The goal of this project is to produce as inexpensive as a possible drone that could help small and medium-sized farmers to have a better view of their land and be able to gather landscape changes autonomously.
There are a few related projects that would be centric to making this happen. Drone mapping software can be costly, but thankfully there’s Open Drone Map project, and its mission is to support the development of an open ecosystem of solutions for collecting, processing, analysing and displaying aerial data.
For the drone to be able to carry out missions, there’s a need for some controller and for the time being, I chose PXFmini from Erle Robotics, a part Acutronics Robotics.
Some inspiration also comes from the blog post on training a drone using OpenAI gym and ROS.
The Heavy Lift Cargo Drone
Heavy lifting drones require a bit more thought at this point, however looking at the progress other people made it seems to be possible to carry a weight of 20kg or more, and this should be more or less sufficient to lift the small greenhouse cube and place it on top of each other.
Some useful information and comparison of heavy lift drones are available here.
Bees are fascinating little creatures, but as many of us heard the bees these days are struggling. There are some great people around the world talking about this problem and trying to get more and more people interested in beekeeping.
Dennis vanEngelsdorp is an assistant professor of entomology at the University of Maryland and also the project director for the Bee Informed Partnership. He has been taking part in many studies aimed at understanding colony collapse disorder. I found his TED Talks presentation — A plea for bees, fascinating. His talk inspired me to learn a bit more about beekeeping, and I should have four bee families later this year!
I want to say a few words about the project that I already mentioned — Open Source Beehives. OSBeehive is on a mission to improve the technology of the beekeeping industry by building a global network of beekeepers and identifying causes and solutions for colony health deterioration. It’s possible to download the CNC files and manufacture a beehive by yourself or at the local maker’s space. The picture above is their Colorado Top Bar beehive.
The Worm House
Figuring out how to convert food waste into natural organic fertiliser is another exciting challenge. The same OSBeehives project offers CNC files for the stackable system where composting worms could live and thrive.
My goal at the moment is to gather a bit of first-hand experience and understand the main challenges. The Urbalive Worm Composter is a kit that can be used indoors or outdoors for composting kitchen waste. I need to do a bit more work this year, but it works, and there are quite a few lessons learned that worth putting into a dedicated blog post.
All these AI-powered robots: the cube, robotic arm, autonomous garden cart, drones, beehive and wormery (with what is the AI in beehive and wormery) is the beginning, and I’d like to think there are more to come. Perhaps directly suggested to the community of this project by the farmers, growers, beekeepers themselves.
Having more and more low cost but artificially intelligent robots available should also allow us to explore the concept of remote gardening. Maybe there’s a piece of land somewhere where some seniors who no longer have the energy to keep the garden thrive. Alternatively, perhaps there is a space of areas where some local data scientists and robotics engineers would like to use to grown their food by focusing on improving the algorithm.
Imagine a farmer or an urban grower could save time if they had a smarter, more automated or a fully autonomous device to grow, process or package the food. A designer or a team could create a design, a software developer or a few of them could write the software, the maker could manufacture such a device in his or hers 3D printer and CNC based small local factory. While using such a method a farmer would produce new data that would share with everyone, and data scientists could find some interesting insights. Machine Learning researches and engineers could build new models and so on. Everybody wins.
How factories manufacture various things is changing too. Instead of making everything on one big centralised factory it is now possible to make more and more high-quality equipment locally. The cost of running a small factory is decreasing too.
Some time ago I discovered a fantastic project called Maslow CNC with its fabulous community around it. Maslow is a community-driven open source project with the goal of making large format CNC technology available to everyone. They believe in a world where people everywhere can collaborate, share, and build amazing things together.
3D printing is also moving from a state where enthusiasts and people working on prototypes use the technology to small scale production lines.
There’s quite a bit of resource about radical transparency online. I’ll mention key areas where radical openness is vital in this project. We want to use the Open Source code and Open Source Hardware design as much as possible. Everything Growbotics AI future community creates would also be all Open Source. If people want to make one of the devices themselves, they can freely do so at no extra cost. If people want to buy something the local maker could manufacture it. We also want to encourage sharing the cost structure for building thing behind the price. There’s nothing wrong with making healthy profits, and profit is good for wellbeing those that work and their families.
When we understand the cost structure of creating and manufacturing various robotic appliances we can all work on helping to reduce the cost without cutting the profits. So when people buy an item produced by the local maker, we want to see the price, the cost of material, the cost of energy, the labour cost and the profit. Moreover, I’m sure the future community that will buy something will be interested in promoting fair and sustainable business.
Another important aspect related to transparency is to produce and use Open Data datasets. This approach should help researched, and curious people around the world work with the generated datasets and find some useful insights. The community working on this project could gather data in many areas: from sensors observing the environment in the miniature hydroponic greenhouse to keeping an eye on the bees or making autonomous garden cart learn how to drive around the farm.
Also, a link to a TED Talks about radical transparency by Ray Dalio — “How to build a company where best ideas win”.
What we often see today when it comes to solar and wind energy is large solar panels and massive wind turbines. Investing in such large projects isn’t what everyone can afford. Often when we see energy plans offered, we’re asked to pay more to support renewable energy. This project should help small and medium-sized farmers to get off the grid. They should be able to produce enough electricity to run their operations. I think an excellent way to address that could be locally manufactured small size wind turbines and DIY solar panels.
There’s an exciting work by John O. Dabiri that is work mentioning. In his “A new approach to wind energy” paper, he compares different designs of wind farms.
So it turns out that 3D printing (and the use of recycled plastic) could help create lower cost electricity. Here’s an example of a 3D printed Vertical Axis Wind Turbine and I’ll be testing one of those pretty soon.
When it comes to solar energy and solar panels, they’re also a lot of exciting developments going on. Many government-sponsored programs encourage people to install solar panels in the roof of their house. However, there are also quite a few man-in-the-middle companies that want to benefit from such move as much as possible.
My primary interest in the solar energy space is to understand how solar panels are made and promote DIY solar panels produced by local makers. Such an approach would allow to reduce the cost of solar energy and also it would help to create different sizes and shapes of solar panels. For example, we’ll need panels to fit the blinds installed on the miniature greenhouse, so solar cells will need to be laser cut. We may also need solar panels to come in the triangle shape to cover the geodesic domes.
There’s some work to be done, but it’s a promising area, and I discovered this great website that can help setting solar panels and monitor your solar energy production.
Making your battery packs is essential too. It’s important for multiple reasons. One is to cut the cost down, but another is to be able to come up with any configuration needed quickly and easily. We’ll need a battery pack for the gardening cart; we’ll need a battery for monitoring beehives. During the night we’ll need to store the energy produced by solar panel throughout the day. Various people already shared information on how to make DIY batteries packs, but I’ll mention one. I’ll mention it because, during a trip to Canada, I met an enthusiastic Belgian tourist who said his dream is to get off the grid and he watches a lot of Youtube videos on how to make this dream come true. He mentioned that there a guy from Florida that talks a lot about DIY battery packs, and perhaps this is the one.
Fine-tuning the thoughts around business model needs a bit more time, but I’ll share some ideas on what I would like to pursue.
There are three vital elements when it comes to the business side of things.
First, this is a commercial Open Source project. Although I believe that many of the outcomes of this project are good for humanity in general, I want people participating in this project to make money.
Second, I’d like this project to be a long term project and not go the route of starting the company and later selling the company to a bigger company.
Third, I want this project to be all about the community. The community that is diverse in all possible ways and has no mental and physical walls and borders.
So how can this community generate income? I believe the simple way would be to launch an online market place and encourage the community members to sell their products via such website. It’s like Amazon, AliExpess, Etsy, and so on but with the focus of small and medium-size farmers selling their fresh products, makers with the help of 3D printing, CNC and small scale manufacturing selling small size wind turbines, robotic arms, electric garden carts and so on.
The small percentage of the sales (let’s say 5%) would go into the shared pot, and the algorithm would distribute it to the contributors: farmers, software engineers, data scientist, robotics engineers, marketing specialists, and everyone who push this project forward.
Also, there’s a great article called “The Building Blocks of Online P2P Marketplaces”.
Another question is how to organise such a community, how to coordinate all the activities? There’s an increasing number of office-less companies appearing, people are talking and successfully trying out a 4-day week. Some scientists, such as Matthew Walker who researches sleep, suggests that 9–5 workday doesn’t suit some people. Talking about sleep, it turns out that it’s hard-wired in our DNA to be an early riser or a night owl. I highly recommend reading his book if you’re interested in your health or the health of your loved ones.
People who believe in the power of decentralisation shared exciting ideas recently. I’ve been a keen attendee of Ethereum London meetup for the last two years or so. It’s a meetup for people interested in the ethereum and other blockchain related projects. One project that I heard about recently is called DAOstack.
However, DAOstack is not the only player in this field. There’s also Aragon, district0x and Colony that perhaps have some similar goals.
Let’s have a quick look at all four of them.
DAOstack — an operating system for collective intelligence. DAOstack powers decentralised companies, funds and markets to make fast and innovative decisions at scale.
Aragon lets you freely organise and collaborate without borders or intermediaries. Create global, bureaucracy-free organisations, companies, and communities.
District0x is a network of decentralised markets and communities. It allows to create, operate and govern. Ethereum, Aragon, and IPFS powers it.
Colony is a platform for open organisations.
All of these projects look very interesting, and a detailed analysis of which one would suit this project best would be needed.
There are still many questions to be answered and explored. For example, would the contribution to the project be only by the people working on it, or where would there be a way to invest in the project? As all the tasks would be listed transparently, maybe someone could be interested in paying extra to make a specific piece of work executed sooner. Also, there’s a question on how to evaluate a particular job, how to measure the quality of the result. Who could suggest new tasks, how they would be prioritised and so on.
It is also crucial to get the community organised in such a way that people can work as much as they want and on what they want. Maybe someone will want to spend one day on this project, and perhaps somebody else would want to spend a year on it. All this effort could be tracked and rewarded proportionally.
So to recap, the future of agriculture could be bright, and Growbotics AI Vision is around these seven principles: Crop diversity, Recycle, AI-powered robotics, Community, Radical transparency, Renewable energy and Innovative business model.
I would also like to add that I attended Yoshua Bengio brilliant talk on “How we can make AI work for humanity” and I’m very excited about what brilliant AI-powered tools we can build to make this planet better for all of us.
Let’s see what we have right now. We have this Growbotics AI blog website, and we have the Growbotics Community website for sharing the ideas, we have Growbotics Market place so that makers and farmer, beekeepers can start selling their products, we also have Github space for all Open Source code. We also have Instagram and Twitter accounts, as well as the Facebook page.
So next will be to keep pushing, make more progress on the developments mentioned above, share progress by posting on Instagram, Twitter, Medium and so on. The blog post about lessons learned while building The Cube should come to the daylight, the results of the detailed comparison of DAOstack, Aragon, District0x and Colony should become available.
Moreover, hopefully, somebody will find this project exciting and will decide to share some of their thoughts.