Today it happened again, the robot found dead on a small treshold next to the charging station. This happens often, but just for a month or so, earlier this did not happen.
If you put the robot in that spot and turn it manually, does it feel like it hooks into the threshold?
Not at all. It just don't know how to get over it... it just drives to treshold slowly, and then trieds to rotate around on it.
Earlier the behaviour was as I thin it should be: Stop before the treshold, then accelerate directly over it. Now it seems that the treshold comes as surprice for the robot.
Ah, I see. Actually, it could be related to the robot being lost or for some other reason having trouble getting to some position. It does some investigation of the neighbouring area - during this movement, it does not correctly look for thresholds. A known bug, sadly.
Looking at the logs, this happened during the robot's return to the charger. It might help if you move the charger to a different spot, further from a threshold.
Today's action: https://youtu.be/QZx9ea3Cuck
First it faced "surprice treshold" but after some rotation decided to do the right thing.
Todays attempt: https://youtu.be/6gfdeQOwq7c
Now it got jammed to the same very low treshold.
It's probably the case that the robot has a hard time seeing the threshold. To test this theory, you could try putting some matte tape on the angled part. Of course not a permanent solution, but could help in correctly identifying what is happening.
Looks that the latest firmware update made a big difference. Now it doesn't run to this treshold anymore like it used to do. No changes in the envinronment, except moving the dock little farther from the treshold, which did not help.
Here is video about one of latest successful runs:
We did figure out an issue with the handling of level differences recently. Happy that it helped with your situation. Thanks for the feedback!
This firmware is definitely better than some previous ones, but I still find it too often either dead on this treshold, or lost to somewhere in the house because of losing location after having problem here.
I tried to reduce the issue by providing more traction by adding anti-slip tape on treshold, and both sides of it.
Having more friction on the treshold looks to be bad idea here, so I removed the tape from the treshold, but left those on both sides.
This guy really tries hard....
After 10 minutes of trying it finally makes the treshold jump maneuver.
Conclusion of this test is that more friction does not help, if the problem is recognition of the treshold.
Could we just draw the treshold to the map in this kind of case to help the robot to avoid such problem?
There is a known bug when the robot ignores thresholds. It's mostly triggered when the robot is lost, I think. I'm affected by it too, so I really hope we get the time to fix it. No known workaround - possibly putting a thin board to make the threshold two steps, so the robot can drive up it without doing its "climbing rush".
Is the behaviour in the beginning of the second video above correct?
In the video the robot is ordered to clean area another side of the treshold: It just drives down the treshold without slowing down. That's few seconds after leaving the charging dock, so it shouldn't be lost at that moment.
In this video the robot also hits the wall quite many times, in very simple environment. Should it avoid hitting objects by using the camera?
Driving down a known level difference is intended behaviour. It is however supposed to do the rush when it's going back, assuming it thinks the threshold is high enough, which it looks to be.
How old is the robot's saved map? There was a bug earlier that corrupted the data in the map about the level differences. If you haven't deleted the map since the last upgrade, please do, to make sure there isn't bad data in the map. This can be done from smart phone app.
Yeah, it's supposed to avoid bumping at all, though sometimes it does it anyway - either because of weird situations or because of it having a hard time seeing the obstacle (shiny paint). Hard to say what is happening without looking at the robot's data.
Map is refreshed after the last firmware update.
Old map is deleted on October 8th and recreated zones after few successful run on October 9th.
Huh, that is weird. I did try the same test and the robot did climb on its way home. I'd recommend you to delete the permap again. But, not sure how it gained the bad link, so not sure if it will come back and how quickly.