Control System Test Needs Updating

[skallaher]

The current control system tests seem to be written for specific parameters. This should probably be remedied.

[hmmwhatsthisdo]

In a similar vein, it appears that the current time-out of 120 seconds may be too short to guarantee the control system test passes - especially because the time-out is based on real time and not the simulator's (dilated) time. The VM on my 810 is currently unable to run tests as a result unless I push the time-out to 180 seconds. (I'm not 100% sure of the reason behind this, but it's possibly a result of the Spectre/Meltdown kernel patches combined with having to do software GPU rendering - I'm noticing a significantly higher proportion of kernel CPU time in htop than I did previously)

I'll most likely open a PR to change the time-out to 3 minutes instead of 2 - there's no particular downside I'm aware of to increasing the time-out.

[irwineffect]

@skallaher, are you referring to the issue @hmmwhatsthisdo is talking about or something else?

[skallaher]

@irwineffect, I was referencing a conversation I had with either you or @ryan-summers about the control system test being optimized for a specific set of PID parameters. I can't remember enough of the details at the moment.

[ryan-summers]

Can we please clarify this issue to say that the overshoot and standard deviations allowed in the control system unit test need to be relaxed? I don't think that has been explicity stated anywhere here.

[irwineffect]

Oh...... we just need to set more sane requirements for the control system test. The goal of the test is simply to ensure that people don't make changes that radically affect the performance of the control system, not to prevent people from trying out different tunings. So we just need to lax the requirements to more reasonable values.

[ryan-summers]

#397