Characterization graphs look off

I’m trying figure what might be an issue that’s causing the this strange graph behavior as the result of using the characterization tool. The Dynamic Velocity test’s velocity tests missing data and Acceleration-Portion Voltage graph.

frc-characterization’s acceleration is computed with a naive second-order central finite difference. The acceleration is basically noise most of the time, and there’s not much that can be done about it. We addressed a lot of data sampling and processing issues in the replacement tool SysId. We’ll be retiring frc-characterization for 2022.

SysId download:
https://frcmaven.wpi.edu/ui/native/release/edu/wpi/first/tools/SysId/2022.1.1-alpha-3
Experimental docs for the new tool:

This tool will also come in the 2022 installer, which wasn’t possible with the old Python tool.

3 Likes

You can trim most of the noise by truncating your data so that there isn’t so much of the steady-state data from the step voltage test. This will help even with the more-robust implementation used in SysId, and iirc there might be a feature there to trim the data?

Yea, you can adjust the test duration to trim out steady-state data.

Thanks, we were able to get path to work. I’ll pull down Sysid and start playing with it. I might be able to try it tomorrow.

The new tool deploys a robot program with more consistent data sampling (so the data you start with is better in general), but you could also use sysid’s built-in JSON converter to analyze existing frc-char JSONs in sysid; the resulting analysis is usually better.

I’ll make it a point to recharacterize tomorrow with Sysid.

This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.