Calibration stuck in RestartingFitness loop

Troubleshooting
1

Edit: I’m now able to calibrate my setup. I have to retract and extend the belts twice for it to work. See my second post.

Hello everyone,

I recently finished assembling my Maslow 4.1 kit and began the installation process. For the frame, I used 4 concrete anchors on my garage floor, and the frame size is roughly 4000mm x 3300mm.

I had read in multiple places on the forum that the orientation of the Maslow doesn’t matter, so I placed it in an arbitrary position on the spoil board and began the calibration using the default values (running firmware 1.0 or 1.01 at that time). The calibration process initially went well—until Maslow ran off the spoil board and onto the garage floor. It turned out the calibration grid was rotated relative to the spoil board. So it seems that orientation does matter after all.

From what I understand, the Y-axis (height) should align with the direction of the router’s power cord. Based on that, I rotated the Maslow 90 degrees to align things properly and restarted the calibration.

However, during the initial belt tensioning phase, Maslow shot right off the spoil board—likely because the interrupted calibration had messed with my Maslow.yaml file. I decided to start fresh by downloading a clean Maslow.yaml from GitHub and also noticed firmware 1.03 was available, so I updated to that.

Since then, I haven’t been able to complete calibration. The process starts, makes the first few moves, and then gets stuck in an infinite RestartingFitness loop. The fitness value rises to about 0.3, then drops back down to 0.02–0.04. The anchor position values being tested don’t make sense and drift away from my actual frame size.

Here’s what I’ve tried:
• Using different grid sizes
• Manually entering the anchor locations in Maslow.yaml before starting calibration
• Letting it run for hours (fitness got stuck at ~0.04 after peaking at 0.3)

Now I’m considering either:
• Changing the orientation of Maslow back to the my first setup
• Rolling back to firmware 1.01, where I had a much better calibration experience

Any ideas what might be going wrong or what else I could try?

Thanks in advance for any help!

Maxime

2

I was finally able to calibrate my setup. After extending/retracting the belts a few times to check the encoders, a 3x3 (300x300mm) calibration ran perfectly. But when I tried a 9x9 (1800x800mm) grid, nothing worked—motors didn’t move, even after a power cycle.

Then I retracted and extended the belts twice, and it worked flawlessly again. This seems to consistently fix the issue.

Usually when I power off the machine, I retract the belts beforehand. When I start the machine again, I often can’t extend the belts, unless I first retract the already retracted belts again. Feels like this might be contributing to the problem. Still testing, but thought I’d share in case others have seen this.

3

At the moment that is the way it works, the machine can’t tell where it is until it does the Retract, extend, tension (or calibration) dance.

4

Lais wrote:

Usually when I power off the machine, I retract the belts beforehand. When I
start the machine again, I often can¢t extend the belts, unless I first
retract the already retracted belts again.

This is expected, after a power cycle, the machine doesn’t know where the belts
are, so you first have to retract them to get them to a known point. YOU know
that they are fully retracted, but the machine doesn’t

David Lang

5

Manual calibration is always an option and worked for me.

6

I understand that, but If I retract and extend only once my calibration systematically fails. I have to retract and extend twice for it to work.

1 Like