FW/VTOL Call: September 27, 2022

📞 Coordination Calls FW/VTOL
1

September 27, 2022

Agenda

First meeting! Compile some topics, see what we can do about it.

Discussion

Fixed-wing runway takeoff

  • Rotate speed vs. takeoff speed (let tecs do it’s job)
    • Consider additionally takeoff climb rate setting
    • Safest is best angle… But we dont always know the performance airspeed for this
    • Max climb rate* would be the second option

@tstastny to make pr

Fixed-wing runway landing

  • Land detection!
  • Cross wind landings need to choose to follow path or follow heading (aligned with runway) during flare

@ryanjAA provides logs for crosswind landings

Y-acceleration feedback for coordinated turns?

  • wish list item from @ben_b
  • Ardupilot has feedback loop
  • PX4 used to have it

@tstastny to look into

Liftdrag plugin

  • Squash and merge

@dagar

Launch detection

Catapult launch: Motor delay not working · Issue #18416 · PX4/PX4-Autopilot · GitHub - catapult motor delay not working

Brought up by @Jakob_Strebel

  • SITL catapult is somewhat working
  • Detects freefall for launch detect
  • Need to check if body-x axis only for catapult detection
  • Shake the plane, log from boot

@tstastny to look into

Failure detection

  • Should not be active in full manual / acro
  • Publish limits from attitude controller

@dagar

Commander takeoff issue (in SITL)

  • Push on commander refactor thread to resolve

Fixed-wing interface clean-up

  • Completely remove actuator_controls topic
    • Split out wheel controller, flap controller, etc
    • Dedicated modules
    • Handling butterfly etc in control allocation (need to config what actuators are used for e.g. “flaps” or “spoilers”)
    • ^use actuator “trims”? (schedule based on “spoiler” and/or “flaps” setpoints)
  • Split out rate controller

@sfuhrer

Fixed-wing guidance

  • purge L1
  • Purge conditionals in aircraft controllers
  • Keep l1 lib for rover

@dagar @tstastny @sfuhrer

2

So in thinking about takeoff…
In the full size work, rotate at Vr, once airborne pitch for best rate (via airspeed) unless trying to clear obstacle then pitch for best angle till obstacle is cleared.

To do this one needs to know speeds for best rate and best angle…
Maybe an autotuning system that finds best rate and best angle?
Being able to basically go out and create the aircraft polar automatically with the real aircraft on a really calm morning would be a really useful thing?

Not sure TECS is the right choice as on takeoff and climb power should be fixed at 100% for best performance. (realizing that a lot of models are grossly overpowered and 100% might be setable to something lower)

3

Hi @pbreed - thanks for taking interest!

once airborne pitch for best rate (via airspeed) unless trying to clear obstacle then pitch for best angle till obstacle is cleared.

Yes - this would be the optimal thing. However, as you mention,

To do this one needs to know speeds for best rate and best angle…

… most folks running PX4 have not identified these performance points for their airframe. In lieu of this, we opted for a mid-term solution of fixing max climb (so the best climb rate, not angle) set by the TECS tuning (which users for sure should have done before trying an auto takeoff!) and at least determining the airspeed they like that max climb rate at. So if someone actually did do a performance calculation, the takeoff airspeed setpoint could indeed be the best climb rate airspeed. Of course this means for now best climb angle is not possible - but could easily be added if more users were interested to take the time to optimize these flight characteristics! I would be fully onboard to try and provide that infrastructure. E.g. one extra climb rate parameter. e.g. FW_TKO_CLMB_RATE.

Maybe an autotuning system that finds best rate and best angle?
Being able to basically go out and create the aircraft polar automatically with the real aircraft on a really calm morning would be a really useful thing?

Yes! There is actually an ongoing effort towards this goal in a university student project supervised by @Jaeyoung-Lim and @sfuhrer. I have high hopes for the result :wink:

Not sure TECS is the right choice as on takeoff and climb power should be fixed at 100% for best performance. (realizing that a lot of models are grossly overpowered and 100% might be setable to something lower)

This I disagree with. I specifically removed the hardcoded “full blast throttle” for takeoff climb out because it was unnecessary and even required some severe pitch minimums in the old setpoint to make sure at full blast the vehicle was actually climbing. – this because TECS was being LIED to / overwritten. TECS will, if tuned, stabilize the airspeed setpoint provided and attempt to hit the climb rate as well. That’s its only job, and if 100% is required, it will command it. But better not to overwrite open loop fixed setpoints on a low level like pitch and throttle when we can avoid it, and have a controller whose purpose is regulating these states.