TUM-Aries-Lab / motor-module / 35220263740
74%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 13
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 72.586%. Remained the same
35220263740

push

github

NGierden 
Fit the velocity slope robustly, and only diagnose a real ceiling

The first bench run on an AK60-6 V3 exposed two flaws in the reporting.

Estimating the scale as a mean of per-step ratios is biased. With kp commanded
as 0 the loop has no integral action, so a constant drag torque produces a
constant velocity deficit -- measured 0.36 rad/s on the bench, consistent to
within 0.06 across a sixfold speed range. That offset dominates the ratio at
low commanded speeds, and the mean over the lower half of the sweep read 612
where the true scale is 812. Fit a least-squares slope instead, which an offset
cannot move, and report the intercept as the droop it is.

The fit also has to stop at the ceiling. Once the command is clamped the
measured speed stops rising, and including those steps dragged a whole-sweep
fit to 488, which then mis-converted the plateau to 24.6 rad/s instead of 15.0.
Fit over the leading run of steps that still tracked, and take the breakdown
point from the same split. The tracking threshold moves from 0.5 to 0.8 because
it now gates the fit rather than only a message, and 0.5 admits a partially
clamped step.

Finally, stop claiming something limited the sweep when it simply never reached
a limit; say so and suggest raising --max-rad instead.

Checked against the real bench data (812.3, no ceiling) and against synthetic
output-shaft-with-clamp (802.1, plateau 14.96 rad/s), motor-shaft and
no-feedback cases.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

1496 of 2061 relevant lines covered (72.59%)

0.73 hits per line

Jobs
ID Job ID Ran Files Coverage
1 35220263740.1 13
72.59
GitHub Action Run
Source Files on build 35220263740