[ZuoCen_Liu]

Thanks,
sorry, Correction, it should be：

liouzuocen@qq.com

[Dhruti]

Sent!

[ZuoCen_Liu]

Small update, Dhruti —
I’ve prepared a v2.4.2 refresh that directly addresses the files and errors you sent:
• Fixed `ValueError`: The crash with `Dispensing_3Axis.csv` was due to a sanitization gap with raw CSV headers and the unsupported 3-axis count. v2.4.2 now handles these gracefully with clear warnings.
• Sparse vs. Dense: I noticed `MainProgram.csv` was a post-processor export. The tool now automatically detects this and flags it as a "Sampling Density Warning" (Screening mode), so you’ll know the result is a low-confidence estimate.
• Full Column Support: It now fully recognizes your `TIME_S` and `X_TCP/Y_TCP/Z_TCP` columns to ensure the highest fidelity for the dense 7-DoF case.
This version should be much more "plug-and-play" for your latest tests.

[ZuoCen_Liu]

Hi @Dhruti,
The dense 7-DoF result, My reading is that the dense export looks clean, which is actually useful: It confirms that SIPA is not simply flagging dense trajectories randomly.
For the next step, I’d suggest a 6-axis Singularity-Skim Benchmark, since RoboDK has a large 6-axis robot library and this would test whether continuity / topology issues also appear outside native 7-axis redundancy cases.
The idea is not to force a bad trajectory, but to create a dense, smooth TCP path that passes close to a wrist-singularity neighborhood.
Even with high-density sampling and a visually smooth TCP path, the wrist axes J4–J6 may show short, high-frequency posture changes near the singularity. SIPA’s associator / continuity analysis could then check whether these hidden peaks are visible in joint space, even when the 3D view looks smooth.

Proposed path:

Position:
x(t) = x0 + L * t
y(t) = y0 + small_amp * sin(t)
z(t) = z0

Orientation:
keep wrist near a singular configuration
add a small ±5° ~ ±15° orientation oscillation

The key is not to enter the singularity directly, but to skim near it.

Suggested parameter range:

Linear path length = 100 mm ~ 300 mm
Lateral oscillation = 5 mm ~ 20 mm
Orientation oscillation = ±5° ~ ±15°
Sampling step = 0.002 s
Duration = 3 s ~ 8 s

It may be useful to create two cases:

B1: safe distance from singularity
B2: closer to singularity

If B1 stays clean while B2 shows localized associator / wrist-continuity peaks under dense export, that would be a strong benchmark for testing 6-axis continuity risk without relying on sparse CSV artifacts.

[Dhruti]

Hi @ZuoCen_Liu,
Let me check all this out and try to get back to you in a day or two!