TUM-Aries-Lab / exosuit-python / 35581641443
88%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 9
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: 87.484% (+0.05%) from 87.435%
35581641443

push

github

NGierden 
Record what the motors report, not just what they were told

Every motor column has held NaN since the CAN switchover: _record_sample
passed not_measured for torque, speed and position. So a session could show a
healthy command stream and say nothing whatever about what the motor did with
it.

That is the blind spot the last bench run fell into. The assist commanded
9.15 rad of output travel on the left leg in bursts up to 5.5 rad/s, with no
saturation and no exception, and the motors did not turn. The recording can
neither confirm nor explain it, because the only motor data in the file is
what we asked for.

Position, velocity and torque are now read for both legs on every recorded
row, in this package's units -- radians and N*m -- rather than the wire's. The
transport's cached feedback is used rather than a request, for the same reason
as the torque read in pre-tensioning: a request blocks, and its frame doubles
as the MIT enable command. Every MIT command already draws a feedback frame,
so while a motor is driven this costs nothing.

The motor's own error code is recorded too. It is one column and it separates
a motor that refused a command from one that never received it, which is
precisely the open question.

The torque columns are renamed from Nm/kg to Nm. They have only ever contained
NaN, so nothing is lost, and what the motor reports is a torque in N*m: on the
MIT path the transport decodes that field against the profile's torque range,
so it is a torque despite being called current_amps. Logging N*m under a Nm/kg
header would have been a unit error waiting to be read off a plot months from
now.

Absent feedback records as NaN rather than zero, so a motor that has never
reported cannot be mistaken for one reporting no motion.

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

685 of 783 relevant lines covered (87.48%)

0.87 hits per line

Coverage Regressions

Lines Coverage ∆ File
59
84.2
0.1% exosuit_python/exosuit.py
1
99.06
0.04% exosuit_python/csv_writer.py
Jobs
ID Job ID Ran Files Coverage
1 35581641443.1 9
87.48
GitHub Action Run
Source Files on build 35581641443