Way more eloquent than my post and nailed it. Previously solved problems.
Caveat -
The S1 has stereo vision for just that purpose.
Turns out, you can’t see when it’s snowing, so that stuff was never actually implemented beyond the “hey, the cams are always whited-out” discovery.
Copy-paste that S1 codebase into a new file called “m1.image” and you have the baseline for the mower.
The order it approaches mowing the no go zone boundaries has little location logic to it at all.
Mine will finish the linear section, pass multiple no go zones on its way to the “no go zone” closest to the areas pathway exit, only to later backtrack to the passed zone leaving unnecessary tracks and wasting battery.
I know that I am not unique with the issues that I have with the routing. But it might help with solutions if I illustrate the specific problems that I have encountered. None of my lawn is rectangular or flat. The pictures below in an examples of some of the mapped areas. These odd shapes are indicative of what I am dealing with. The mower does a decent job, but the mapping is difficult and routing is ridiculous.
When I did the mapping shown in figure #1, the little jog at point “A” wasn’t there. It didn’t show up until I filled in all of the parameters for mowing, and then did the preview. Probably a GPS blip, but annoying as hell as it causes the mower to stop and start a lot and chew up the lawn. I know that I can try to modify the map by driving Yarbo back over to clean up this area, but I had no luck connecting the new line cleanly. I just ended up with a blip somewhere else. For something like this it would be really nice if in the app I could just delete the errant point to smooth out the path. Redrawing with the unit is way too imprecise for something like this. So I get a lot more wear-and-tear on both the Yarbo and the lawn.
The routing problems are a much bigger deal. First, the pathway to this work area enters at point “B”. The routing algorithm decides to start the mowing about 60’ across the lawn at point “C”. My back yard is even worse. The path enters at the top end, and the starting point is at the entire opposite end of the area. I travels about 120’ across the lawn to start the process, and of course ends there as well and has to travel 120’ back, trampling the grass in both directions.
And because of the odd shapes of these areas, I have found that the spiral pattern is a much more efficient pattern. I get somewhere around half as many turns as I do using parallel. But there are these weird little anomalies that make no sense. It does 95% in a good spiral pattern, but then it leaves the center of these various spirals for last. You can see the four areas marked in green that are done last, which requires much more wear-and-tear on both the Yarbo and the lawn. Why??? Please tell me how this makes sense. I am willing to learn.
One last note about routing. Unlike a lot of people, it seems, I don’t care at all about stripes. All I want is for the lawn to be mowed completely and efficiently. Minimal abuse of the grass and machine. So, for a work area like figure #2, I would prefer that the mower just keep on going and mow over areas that have already been mowed instead of doing what amounts to a total of 9 turns in a very small area that is then totally destroyed by the tracks. I tried setting this area up as a long looping pathway or a bunch of dead ends, but it was too difficult to see exactly where I had already been and to drive straight enough paths, so I gave up. At this point, I just mow this area myself.
Let me end with stating that I love it as a snow blower (not that it couldn’t be better), and it isn’t bad at mowing. If these problems could be solved, I would be able to say that I love it as a mower as well.
As far as where the mowing starts I have heard others say you can influence where the mowing starts by placing a very small no go zone on the perimeter. Even a tip of a triangle that is mostly outside the parameter seems to start the mowing there. I have not tried it, but it might be worth a try.
My experience is similar to yours with spiral. Most of it is good, but the middle parts can get messed up.
I have a long skinny area of grass next to my driveway that spiral works best on, but I too get weird zig zags when it gets to the middle.
I’d be ok with spiral mowing area’s its already mowed to help clean up those zig zag strips.
Like your figure 1 area 2. Come around the no go zone towards the top of the image and pass through the number 2 area. Might be a lot more traffic between the no go zone and the perimeter. We might end up not liking that solution either.
“dynamic overlapping” allow it to overlap more to make the future paths more efficient! Some areas it really needs to just mow forwards and backwards and get completed first or save for last, to make the rest of the area more efficient…I think Adaptive somewhat does this but still not super great. Today I even did a double take as Yarbo was “facing the wrong direction” but it just mapped a little triangle…would have been 10x faster to just back up and turn a bit to get to that corner😅
Yarbo’s mowing can be much more efficient. It doesn’t seem to realize how big it is. Worst areas tend to be “bump outs” where it spends so much time spinning and back-and-forthing in an area just 2 Yarbos wide. It could literally go in, back out, in, back out, done.
Thankfully the efficiency has improved somewhat, hopefully it will get even better over time.
Definitely has improved and I know it’s on the priority list. Hopefully the short list.
Yeah I have a bunch of small trees and such near the back of our property, I tried mapping with really small no go zones so that it could try routing more efficiently but now I’m considering just making bigger no-go areas and then putting dead ends in…seems like a lot of extra work that really shouldn’t be needed…I’m hopeful that by next mowing season they get a lot of this stuff figured out, the PPVS update wasn’t perfect but very encouraging and hearing about the VIO work being done is also great! I also follow Tesla self driving and knowing the hardware similarities they definitely can be epic if they get the software figured out!


