The first thing to do when starting off with a new idea or a new project is to Google it.
First of all it forces me to narrow down what I am actually trying to create. A Google search is only as good as the search terms, so having a vague description is not ideal. I usually start with something vague, realize that the internet can't read my mind, and force myself to take a step back and narrow down what I am trying to figure out.
Secondly, there are no new ideas - Mark Twain said it so it's gotta be true. Most of the time whatever I think I have come up with, someone has come up with, tried it, and either failed or succeeded. If they succeeded - great! I can use what they created to learn how to do the thing. If they failed - great! I can find out where they failed and not repeat their mistakes, or even better realize that it's not going to work for a fundamental reason and move on. Fail Faster!
So lets start and see what we get.
"high tech backyard sustainability"
Ok, pretty vague but it hits the general idea I'm after here. There's a lot of noise in the results. A lot of articles on yupsters building 800k eco homes - not really what I'm after.
However further down I found an article about FarmBot: https://farmbot.io/
Now this is more like it. an open source gardening CNC robot with computer vision weed detection and a web app interface.
A complete kit is $2,500 which makes me wonder how well this would do in the "money is money" metric.
The Farmbot site has an ROI breakdown where they calculate out the money that you would spend on the Farmbot kit, shipping, and installation would pay itself off in 2.5 years from the amount of money saved by not buying veggies at the grocery store.
That's actually a good way of looking at it. I wonder how many other systems can be calculated out this way - in terms of how much time it will take to pay itself off.
The robot itself is based off a CNC machine and the kit looks like it uses some high quality parts. Potentially too high quality - the tolerances of planting a cabbage seed as a bit looser than milling. I feel like I could build a similar system using cheapo stepper motors and aluminum rails.
Their kit also uses an interchangeable tool head which can be used to plant seeds, water the garden, test the soil, elimite weeds, and hold a camera. The only head that seems necessary is the weed remover - as that would be the only tool that is used continually.
I can plant the seeds - or starts more likely - by hand. I can have a drip irrigation system connected to a control valve and use permanent soil sensors to read moisture and perhaps ph. The camera could be permanently attached to the gantry. I appreciate the "all in one" approach, but the added complexity - both in terms of construction and maintenance - could make it less efficient in the end.
So then, onto the software. The Farmbot uses a Raspberry Pi 3 as its controller and a web server app to actually program and control the robot. The web app looks like it has a lot of features - planning garden layout, directly controlling the robot, scheduling events, detecting weeds. To be honest the only thing that I really want is the weed detection algorithm but it will probably be easier to use their software stack as is then trying to pull out the pieces I need. In that case I will have to use a RasPi 3 and whatever steppers they have in their kit - or at the very least the same stepper driver.
So I guess that's a plan - go through the farmbot documentation, figure out if I can build a cheaper version, and see if I can run it using their software stack. Also now I have an idea for the next search term: "computer vision weed detection" or "computer vision plant health". Guess we will see what that leads to tomorrow.