Use gyro/accel/mag raw values for SCALED_IMU mavlink msg

[ChristophTobler]

We currently need to stream the raw/unfiltered gyro/accel values on the Snapdragon via Mavlink for VIO estimation (probably also on other platforms). Since the raw values don't seem to be used (was even inconsistent) I changed to to be the unfiltered values. There can then be streamed via the SCALED_IMU Mavlink msg.

Hopefully, I got all drivers...

[mhkabir]

Just curious - what's wrong with the values in HIGHRES_IMU? I've been using it for VIO for over a year now without any issues.

[ChristophTobler]

@mhkabir For two reasons:

• Filtering adds a delay which is unwanted
• Most VIOs use a EKF which already is a filter so it doesn't make much sense to prefilter it

We've tried using HIGHRES_IMU without any success on the Snapdragon with VISLAM. My assumption is VISLAM is very sensitive to delay/timestamps.

[LorenzMeier]

this looks like a valid short-term solution but the changes to all sensors are incomplete. I'm still pondering if this is the right incremental step or if we need something different (unfiltered, but correctly scaled values) across the board and rather should be doing that.

The current scaling does not have enough resolution to not introduce integration errors, which means its suboptimal for VIO and not viable long-term.

[ChristophTobler]

@LorenzMeier I added the missing drivers (hopefully, I got them all).

I was wondering about the resolution as well. But what what do you mean with integration errors? Rounding errors? The thing is, it works reliably with VISLAM and therefore I'm not too worried.
Otherwise we would need to change the uorb msg or add a different one.

[LorenzMeier]

@ChristophTobler The integration errors would not be visible in terms of diverging, but in terms of the convergence in EKF2 showing a constant error and the filter never converging fully to the right offset.

[ChristophTobler]

Closing in favor of #8591 (temp hack) and new issue #8592

[LorenzMeier]

We are moving towards this interface for onboard companion communication: https://dev.px4.io/en/middleware/micrortps.html

[mcsauder]

My apologies @ChristophTobler, I didn't mean to post that comment here... not sure how I managed to accomplish that. I'll delete it now so that it doesn't mislead future readers... Again, my apology.