FW Position control: switch from L1 to NPFG guidance by default

[sfuhrer]

Now with the next release branched off I think it would be a good moment to start transitioning from L1 to the NPFG fixed-wing guidance. .
It will give us more robustness in wind, deterministic flight paths and cleaner interfaces.
As a first step I would propose to make it default but keep the L1 option, and get more testing feedback in prior to removing the legacy L1 controller.

[tstastny]

I see no downside to this (and keeping both is becoming a bit of a headache in the code) - but probably we should hear from others who weren't the one who wrote the new library in question @dagar @RomanBapst @Jaeyoung-Lim

[Jaeyoung-Lim]

This will change the behavior of all previous fixedwing platforms if the existing planes have been tuned for L1.

While I agree that there is no harm on making it default for SITL vehicles, shouldn't we at least keep the existing air frames(real airframes) to keep using L1 for real vehicles unless it has been flight tested with NPFG?

If they are not maintained, we might as well delete them

[tstastny]

This will change the behavior of all previous fixedwing platforms if the existing planes have been tuned for L1.

this is not actually true. the period and damping params are one-to-one reusable. only in very high winds does the behavior change.

[dagar]

I see no downside to this (and keeping both is becoming a bit of a headache in the code) - but probably we should hear from others who weren't the one who wrote the new library in question @dagar @RomanBapst @Jaeyoung-Lim

Unless we want to look at thoroughly testing both on an ongoing basis I think we should figure out a full (safe) transition plan and fully drop L1 once we're confident.

[tstastny]

a full (safe) transition plan

@dagar should I start an issue to populate some objectives? e.g. (off the top of my head):

• how to transfer tuning of existing systems
• SITL (functional) tests for various waypoint switching behaviors (there have been a couple hiccups in the last months, as typically expected. Just had to add a condition to navigator.. but these are things that need to be flushed out)
• flight testing bare minimums before transition (hours? platforms? modes?)

or would here and/or another place be better to organize this?

[Jaeyoung-Lim]

this is not actually true. the period and damping params are one-to-one reusable. only in very high winds does the behavior change.

@tstastny Right, but isnt the parameter name different and if you have a custom period, it will roll back to the default with this change?

[sfuhrer]

@tstastny Right, but isnt the parameter name different and if you have a custom period, it will roll back to the default with this change?

We could add them to the param translation. Though I wonder if that doesn't let to worse behavior than setting them to the new default (10), given that the old default was 20. @tstastny why again is there so much room between the old and new default? Did the old one have to be that big bc the controller itself was less robust?

[tstastny]

Npfg has the optional functionality to automatically increase the period when the condition and current tuning would result in unstable behavior (assuming one has properly identified the roll time constant). So a lower period can be set without worrying. -- as to why 10s specifically, that was my preferred tuning for an easyglider. But has also been used on larger aircraft. 20s for l1 was probably set before as you mention to be a safe default for most aircraft.

We could bump to 20 for default if folks felt safer.