@ronlawrence3 had an idea to make a standalone site that simulates calibration (similar to his calibration tab), so I went ahead and worked on making one! You can put in the CLBM line from your logs (or an edited version of your own) to run the calibration again.
Behold:
Maslow 4 Standalone Calibration Simulator
Repository is here: GitHub - etskinner/Maslow4-Standalone-Calibration: A standalone site to simulate calibration on Maslow 4
So cool!!
Let me know if you find any insights into cases where it isn’t converging
@ethan or @ronlawrence3 , or anyone really - I’d like to play around with this but I don’t have a Maslow4 yet. Could someone post a few examples of input data, especially sets that seem to have been collected properly (no slack belts etc.) but don’t converge to good fitness values.
I’d like to do some playing around and analysis of the data.
Thanks Ethan!
Sure here is my latest 4x4 calibration at 400x400mm (also see calibration failure topics for others’ files)
CLBM:[{bl:1701.36, br:1723.27, tr:1732.32, tl:1760.84},{bl:1576.98, br:1598.84, tr:1865.81, tl:1892.58},{bl:1419.47, br:1775.27, tr:2013.68, tl:1753.69},{bl:1511.88, br:1847.81, tr:1930.20, tl:1657.11},{bl:1604.85, br:1923.37, tr:1852.66, tl:1565.97},{bl:1698.12, br:2001.57, tr:1781.73, tl:1481.16},{bl:1784.07, br:1896.97, tr:1668.29, tl:1588.32},{bl:1694.99, br:1815.02, tr:1743.91, tl:1667.53},{bl:1606.75, br:1734.76, tr:1826.08, tl:1753.52},{bl:1521.40, br:1657.76, tr:1914.05, tl:1844.97},{bl:1632.50, br:1544.97, tr:1818.70, tl:1941.19},{bl:1711.99, br:1628.13, tr:1725.90, tl:1854.43},{bl:1794.19, br:1713.46, tr:1638.74, tl:1773.39},{bl:1878.47, br:1800.53, tr:1558.14, tl:1699.03},{bl:1982.04, br:1712.76, tr:1451.70, tl:1812.93},{bl:1900.98, br:1734.00, tr:1537.91, tl:1882.58},{bl:1823.67, br:1527.87, tr:1630.52, tl:1959.27},{bl:1748.59, br:1441.69, tr:1728.46, tl:2041.48},]
I’ve got a version up on @Ethan 's repository here that will allow you to select entries to run various scenarios with. After pasting in your data, you can run the simulation with any/all of the measurements and interrupt it with “Stop” and it will print the progress on the bottom.
Merged!
New version for 0.68.1 with initial / best guess manipulation. it will update the 5 values as the calibration calculations run.
Site:
https://m4cal.etskinner.com/
Screenshot:
Merged! FYI Ron, since I added you as a collaborator, I think you can merge them yourself. Let me know if it doesn’t work.
Will do next time, thanks!
@bar watching the values change is interesting. I notice that the points move even when giving the same fitness. I don’t think its my code causing that, but I found it somewhat surprising.
I have a branch of this with the 0.70 functions in it. I need to fix a few things before I publish / merge, but the basic calculation works ok.
https://deploy-preview-3--stupendous-sherbet-f63d82.netlify.app/
Nice work!
I just tweaked the algorithm to make more aggressive about moving towards the goal and it’s it 10x faster and it seems like it’s actually more precise too.
I’d love your feedback if on if that change seems to be stable. I found that if I made it too aggressive it could become unstable and fall apart
Great, I’ll pull it in and compare. The thing I need to incorporate here is the easy setting of the guess data from the console of the maslow. Plusses and minuses to having the ui separate from the data 
I have the above link updated with the latest code. I still need to get to the entry of initial values from the maslow, but I’m about to step out for the night, so I’ll work on it later or this weekend.
And… I am converging on the same answer here I just got from 0.70 without initial values
