@bar started a github issue to have the system automatically figure out the calibration grid size and number of points to measure. I jumped in and directed the AI to clean it up a bit.
It’s now ready for testing. It needs to be tested on different frame sizes, and it defaults to positioning the measurement points up to 400mm apart (configurable in maslow.yaml)
I’m willing to give it a go, uploaded the firmware. Ran a Find the anchors. All appeared to be going well, until the Maslow headed for a dive off the table. I was half expecting that so I was ready and shut off power to the M4.
Problem: the size of the automatically generated grid.
X not a problem but table is 2400 x 1200mm. Suggest we still need to set a max height and width parameters for workspace, especially when anchor points are extended like mine.
download the firmware-package file, unzip it and you should have firmware.bin
that you can load
set the grid size to 0 x and 0 y and then when you run a calibration, it should
pick the size and pick points to use within that (minimum of 3 in each
direction, by default, 400mm spacing)
keep track of the number of points it picks, how big an area it uses (I’m not
sure how much detail it puts in the logs, something I’ll look at soon)
but the real question is how good the anchor points that it picks are compared
to what you currently have from the old process and the resulting fitness score
Hi David,
I managed to work that out eventually, loaded the firmware and ran. I then set the dimensions to 2400 x 1200mm for the work surface and ran it again. It ignored what I had put and headed off the table at the same place. i didn’t save the log files the second time. It actually got to the same place much quicker the 2nd run with the values in the yaml file. I will compare the anchor points and fitness score, rememering it has not completed a run due to me having to stop it.
I managed to work that out eventually, loaded the firmware and ran. I then set
the dimensions to 2400 x 1200mm for the work surface and ran it again. It
ignored what I had put and headed off the table at the same place.
that’s a good point, we haven’t had it check the work area size, it’s just been
checking against the frame size.
what is the frame size again? (roughtly)
i didn¢t
save the log files the second time. It actually got to the same place much
quicker the 2nd run with the values in the yaml file. I will compare the
anchor points and fitness score, rememering it has not completed a run due to
me having to stop it.
I think this is going to use fewer points by default with the current setting
Ok, a new one to try, this one adds a sanity check for wide, but short frames
it also adds a block of output to tell us what it’s doing.
something along the lines of:
Calibration area sanity check: angle at top = 115.8 degrees (should be < 140)
=== Calibration Grid Configuration ===
Grid dimensions: 909.6mm × 909.6mm
Calibration area: 0.8 m² (827437.3 mm²)
Grid size: 9×9 = 81 points
Frame dimensions: 2438.4mm × 2438.4mm
Clearance from edges: 764.4mm (width), 764.4mm (height)
4.046m x 4.674m measured with a laser. approx 160mm longer TL → TR length BL → BR
Diagional 6046 x 6173 mm
Thanks, that’s a big frame. it needs ~14.2 of the 14.5’ of belt, so be very
careful in the corners, some people have reported having the belts pull off the
spools when they got too close to being fully extended.
I’ve got a new firmware blob that limits the max calibration area to 8’ wide and
4’ high
@bar 's first cut was to set the calibration area to 50% of frame size (just
over 2.2m x 2m in your case). I refined that to the shape of the frame, but move
a bit inside the green (about 2m x 1.8m square in your case. both of which would
have gone way outside the area of your spoilboard
This latest version stays within a 4x8’ working area, which may still not be
enough, so watch it.
Not good, same again, M4 attempting suicide at the same cliff.
hmm, the AI is saying that wasn’t the version of firmware that had the 4x8 limit
in it, try this one
I don’t know if I sent you the wrong link, or you confused the two (I’m looking
forward to the enhanced versioning getting merged, I will admit that I got a
thrill of seeing your maslow check for a new version, even with autoupdate
disabled
Sorry thought you must have given up for the night. Will give this one a go, but I’m pretty sure I loaded the latest version. Offsets are 10mm, wall mounts are pretty close to level, an earlier version had a problem (probably divide by zero when I set them to 0)
