Top threat to every lawn bot, ever:
A moving car on a driveway.
Try to avoid ever crossing a driveway. If you must, then you must. Otherwise, go completely around the house if you have to. You think your spouse will watch for the robot, and they will. They’ll see it across the lawn and put it in reverse, and then they’ll hear it crunch under the car. Or the FedEx guy, or Amazon guy, whatever. Every lawnbot has a car detector, and the first thing the bot does is drop everything and head directly toward it. Yarbo’s collision avoidance looks “down”, not “up”. And that applies to everything.
If you are trying to map an area that loses GPS, just skip that dead zone area. Map the rest, tune that, and come back later. You can try to add more of that dead zone piecemeal, and prove the robot works reliably in that addition. That’ll put the robot on the edge of what can and cannot work, instead of sending it 200 feet into the void.
Do not try to be accurate around hazards. Give them a wide berth. You tighten things up once you learn the behaviors and caveats of each specific area and mowing angle. Mowing a plot with a 35 degree angle can be completely different from mowing it at a 70 degree angle. And it changes, yet again, when the morning sun is blinding the cameras, vs mowing in the afternoon. Vetting the behavior of the bot in an area takes time.
There’s an instinct to get an area’s plan perfect on the first try.
Ignore it.
Protect the robot from things in the yard, then expand the robot’s area once you see where and how it screws up, vs where it doesn’t. It takes time to do that.
These things are awesome, but become less so when it decides to bushwhack through the woods, take out your kid’s play set, suicide under some stairs, or drive into a 12 foot ditch. If you pay out your areas carefully, you can probably avoid most of that. As long as YOU avoid such things, you’ll be smiling. If you expect it to start threading needles, you probably won’t be.
