RTK GPS is such a cool technology. Maybe it’s because my undergrad was filled with inaccurate data but turning on a unit and having 1cm accuracy still feels like magic.
Just deployed a Segway Navimow which uses RTK + vision for navigation.
Going from Landroid (wired, fails frequently) to Husqvarna Automower (wired, fairly reliable) to now a GPS RTK solution is quite the leap in technology.
Husqvarna does have an RTK unit but it is quite pricey ($6K). The Navimow is surprisingly inexpensive ($999 for entry level unit). The Mammotion Luba 2 is sort of in-between.
The RTK units are only becoming widely available in the last year or so.
I just so happened to set it up on solar flare day so it's hard for me to say how much that impacted it. It does seem to have a much more consistent signal today, though.
It's quite something to see it precisely drive from one end of the garden to the other, especially after years of managing wires.
I would say the RTK is quite good and accessible at this point. It does require placing the base unit in a place with both a good view of as much sky as possible, as well as being able to communicate on 900mhz to the bot. The bot itself receives GPS signals as well as the 900mhz. The tower and the bot have to see ~5+ of the same satellites at the same time. This is why it's key for the base station to have a good sky view --the bot will constantly be seeing a different subset of satellites, so you want the base station to see "all of them" if possible.
My dream outdoor bot would use a combo of RTK, lidar, vision, as well as WiFi positioning techniques.
I wonder how accurate that could get?
Maybe overkill (RTK alone seems "good enough," but there's still a gap between precision of these outdoor bots and, say, lidar-based indoor vacuum robots.
Oh yeah, the dream bot would also use photogrammetry plus all the other data to reconstruct a full detailed mesh of the outside.
I want a push mower because the available robomowers all have tiny razor blades instead of a actual mower blade and a limited range. My lawn isn’t gigantic but it’s not perfectly smooth or even and I dread paying that price for something that either won’t work or frequently breaks the blades.
As far as lidar and wifi positioning - that could give you sub-cm, but I would go away from positioning to create a mower that knows the difference between a road and a lawn, a flower and a weed, and your lawn and your neighbors. A dream would be just giving the mower the same ‘mow the lawn’ command you would give a teenager and it just working. With the way things are now, these mowers have to be driven around the perimeter and no-go zones created manually. I saw one guy whose lawn was more of a field plotting out the entire mowing track in software with his automated push mower. It’s better in the software then going out to the field to drive it like an RC car, but it’s still far from some5ing like ‘full self mowing’
> the available robomowers all have tiny razor blades
I was concerned about this as well when I deployed the first robot.
After having a bunch of stumps removed from my yard, all my wires were messed up. So I actually switched to a big zero turn mower for about a year.
You know what I found? The cut quality was worse, even with a sharp blade and perfectly leveled deck.
The powerful traditional rotary mowers kick up a ton of dust, the wheels are much higher impact on the grass, and being a manual process it isn't practical to do every day or multiple times a week.
We realized that we actually liked the "robot grass" much better, so at that point I fixed the wires and now the yard is better than ever.
The reason the robot grass is much better is because it can cut much more frequently. I have them going daily in the AM and the yard is perpetually freshly mowed like a golf course. I just string trim about every other week.
I'm not worried at all about the little blades (and we have a fairly rough yard being in a clearing on a mostly wooded lot on a hill).
They recommend replacing them every month, but I get good results only changing them once a year.
I deployed a fully autonomous automower a few years ago which navigates by sight, sound and terrain-following sensors. It is guided by visual cues which tell it within which area it is supposed to keep the vegetation short. It does not need to be guided any further since it senses where the vegetation is in need of clipping. One of the best things about this system is that it does not require charging since it actually runs on the clippings it harvests. The remarkable thing about this solution is that it is not new, the current incarnation is 33 years old.
It has self-reproduced twice.
It is a small to medium sized Welsh mountain pony, keeper of grounds here at our farm.
This model has the design quirk that it concentrates its waste in a specific spot which negates the problem of stepping in fertiliser but which does mean it is up to us to spread it. We tried other models which did not show this behaviour but which ended up being too heavy for our type of landscape - a hilly, silt/clay mixture which does not provide enough support for heavier models when wet - so we'll use this one until it expires.
About 2 acres. Right now there is a Husqvarna 315 and 415x doing the big open spaces. I have the i110n doing all the small/random/scattered areas that are hard or impossible to do with a wire.
That said I think the whole job could be done with one single RTK mower that is rated for a bigger space, but that wasn't availa available at the time and I've been evolving the setup over the years.
Years ago I experimented with this really cheap GPS. Pixi or something like that. It had poor ergonomics and as difficult to establish a connection. But once it worked it was accurate. Was like 1/10th the cost.
Also there’s VRS RTK without a base station (but you need internet and a subscription) for some reasons. Works just as well. It’s basically turn key.
Yeah, I remember the Pixi (sp?) kickstarter, back in the heyday of ArduCopter. There was a brief period there where open-source, open-hardware, DIY got you the best results for multicopters.
