Dev Call Mar 04 2019 2300 UTC

Issues & Pull Requests


  • plane 3.9.6 released



Attendee count (max): 23

UTC2302 - add tailsitter input roll/angle pitch euler method

  • Mark is just retesting in RealFlight

  • Tridge has made a small review with minor things

  • New mode, so Do No Harm should be in play

  • Can be merged after the minor things are merged

UTC2305 - QACRO for tailsitters

  • Tridge thinks this looks good

  • Will need a follow-up PR to add QACRO to pymavlink

  • Should add to the mavlink PLANE_MODE mapping in the mavlink

  • Restriction to not enter this mode when not a tailsitter is in place

    • Tridge thinks we should allow it in ordinary quadplanes

      • Mark has tested and it works OK on ordinary quadplanes

      • So we will drop that commit

        • But add a warning that it is not really suitable in docs
      • There are some very sporty quadplanes out there

  • merged

UTC2312 -

  • Need to reduce gains for high-speed travel

  • Control surface oscillations otherwise

  • Mark has flow his Dart at 18m/s with this patch

  • Should be using speed-scaling from Plane with an airspeed sensor?

    • Steps in the speed scaling have been a problem recently

      • Jitter on the ground

      • Close to zero velocity in speed scaling code

      • Need to make sure things are continuous

  • Can be turned on and off with parameter

  • Wind estimate problems were bad during transition from vtol to fixed wing

    • Need to do a rotation in the wind estimation code

      • Assuming velocity in x/y plane, perhaps?
    • DCM wind estimate is unlikely to be good while hovering

  • Behaves differently with/without airspeed sensor

  • Speed along axis of pitot tube is what we need in our simulations?

    • Is that what we’re actually getting from FlightAxis?

    • Tridge fixed something along these lines in normal SITL recently

    • [10:20 AM] (Channel) MdB: built in quadplane sim doesn’t handle that correctly for wind when hovering

  • Has this been flown on real vehicles?

    • Some good feedback/videos

    • Iampete has run this and is working with Mark on what he’s seeing

  • Iampete is keen to move on with this PR on scaling

  • This is probably better than what we have now

  • Can pull pete’s method in later as an improvement or option

  • Merged!

  • Zero airspeed when in hover in SITL if you are not moving even if you have wind simulation on

    • There is an open issue on it

UTC2325 -

  • Subclass for tailsitter quadplanes

  • Randy should be happy with the approach

    • He hasn’t looked at it yet
  • Keeps existing tailsitter support

    • Doesn’t touch class-10 tailsitter at all

    • Won’t break existing setups

    • Adds a matrix-based tailsitter

    • If a parameters is non-zero it instantiates the new one

    • Why not an M_CLASS?

      • Uses same class as quadplane does

        • For quad-hex etc
  • Waiting for Randy to OK it

    • Does it fit the architecture philosophy or does he want to go in a different direction for this sort of thing

    • Do we need a motors library?

      • Consolidate the outputs to the servos rather than having them scattered around

      • Like in Rover

        • Takes high-level commands like roll-left or this-much-throttle

          • Motors lib turns those into servo outputs
      • Tridge: Want to take advantage of AP_Motors in quadplane

        • Has to take into account all of Plane’s capabilities

          • And all of AP_Motors

          • Instantiating AP_Motors is based on vehicle type?

            • Rover isn’t like that

            • Rover internally decides what sort of vehicle it is and does stuff itself

              • See video of ackerman + skid steering vehicle
          • Huge combination of vehicles

      • Interfaces stop letting people do weird things all over the place in the code

      • [10:37 AM] To Weekly devcall: … merging Tom’s stuff would be a useful precursor, shall we say? :slight_smile:

  • Should be harmless as it doesn’t seem to change anything existing

  • This is similar to the way tricopter stuff was done

UTC2337 -

  • Merged!

UTC2339 -

  • Interpretation of quadplane in Plane is complicated by rotations

  • Move responsibility up to base class

UTC2341 -

  • Less OBC-specificness?

    • If rules changed we’d change this

      • Or maybe add a OBC2020?
    • Red if armed, green if disarmed

      • Red is dangerous!
    • Randy likes the OBC-specificness

    • Docs don’t say what OBC does

    • Change description to say what OBC does

      • Green for safe-to-approach

      • Red for not

    • More schemes?

      • Scripting?

        • Safe-to-approach case is critical, scripting can break?
      • What’s the threshold for adding new schemes?

    • Nate thinks something agnostic

      • E.g. scheme1

      • Randy Name vs number makes a lot of sense

        • Can search on it
    • SafeToApproach rather than OBC?

    • What would MdB need to add his in?

      • Tridge: we should have more schemes

      • Put this in AND Michael’s concept

  • ProfiLEDs?

  • serial-LEDs

    • We now have DMA

    • DMA series of pulses out

    • Tight on timing

      • Same as d-shot
  • Link to OBC rules?

    • Or just describe it
  • No long descriptions in values

  • Alphabetical order for descriptions on long lists

    • [10:52 AM] (Channel) MdB: So description has to be sorted by us not a GCS

UTC2351 -

  • 3 weeks away ‘til rollover

  • MdB doesn’t think there’s going to be an issue

  • [10:53 AM] (Channel) JW: It’s Y2K all over again

  • [10:54 AM] To Weekly devcall: Except we really HAVEN’T done all that hard work that went in in the years up to 2000 :slight_smile:

  • We don’t have valid GPS week for everything anyway

    • Except for UTC feature

    • In which case we already had the problem

    • Absolute date to trigger command might not be doing well

      • NMEA doesn’t allow it anyway
    • UTC Time can already go backwards anyway (up to ~12 minutes after boot….)

  • Convince ublox to go forward in time?

    • Simulator would do that

      • US$5000
  • Really old GPSs (<2002) will probably have problems

  • No problems with GLONASS

  • Consensus is we won’t have issues

  • [10:59 AM] (Channel) RM: MdB, if you have any comments could you add to that issue? I’d like it if I’m not the only one on there…

UTC2357 -

  • Adding crow flaps option parameter

  • Bitmask so we can do a few different things with crow spoilers

  • How does this interact with differential spoiler?

    • Four different servos inner/outer and left/right

    • This uses that

    • By default not change

  • By default the old dspoiler stuff just used elevons

    • Pitch on spoilers as well as roll

    • Can now say only-roll

    • Turning on/off different axes of control

      • Bitmask should talk about roll/pitch
    • “Progressive crow”?

      • Currently flaps are dspoilers and ailerons go up

      • This does flaps-down first then 50-100 ailerons go up

      • For really throwing the brakes on

  • Minor things

    • Default constant of three should use the bitmasks

    • Shouldn’t be calling things after vehicles

    • Not “flying” wing but “pitch control”

      • Elevons vs ailerons

      • Random flying monstrosities make “flying wing” a non-representative name

        • Axes are better
  • Testing using “marlin” in RealFlight

    • Add parameters in SITL models repo?

      • So tridge can test this easily
  • Should be able to make gliders much better

    • The ones with no motors….

UTC1107 -

  • Prearm check inside controller rather than just checking PIDs

  • Peter to redo as a general prearm check on the controllers

UTC1108 -

UTC1113 -

  • Make it a waf configure option?

  • Made it so much easier with SkyRocket

  • FF point that date doesn’t tell you the code is correct

  • But normal users like dates

  • Also, we’ve only just made changes to get consistent build dates

  • So configure option to turn the dates on

    • –with build-dates
  • Not on by default

UTC1117 -

  • Any discussion?

  • Compass learn works with this

  • We have a SITL mode for a vehicle that spins around

  • This can go in once compass-cal is seen as working

UTC1120 - target system field

  • Can go in straight away

  • Hash on the field name so we can’t fix the field name

UTC1121 - can we merge mavlink pull request #90?

  • Merged already!

  • Tested against ArduPilot master

UTC1121 - Plane 3.9.6 is out

  • Thread creation fixed on Matek405 wing

    • Crash on boot

    • Just pipped over threshold for fitting static data for threads in main memory

      • So tridge shifted into spare 64k secondary area

      • Thread creation via ArduPilot API rather than ChibiOS API (which can only use main heap)

  • Tridge is thinking about 3.10 sooner rather than later

    • Lots of good features in 3.10

    • Copter plans?

      • Not really ATM
    • “New stable” and “old stable”?

  • Some reports of parameter resets on boards using flash for parameters

    • As opposed to FRAM

    • Need a reproducible test case!

    • Could be a HW problem?

UTC1123 - Randy and Rover


  • 3.6.7rc1 is out

    • Started late last week

    • Early-landing fix

    • Compass fix

    • Want to release in next couple of days

    • Zubax GPS delay time

      • Lost the delay time somehow?

      • Want to bring up devices post-boot!

    • Not too many big issues in last week

    • Compass calibration not making noise when kicked off with transmitter

    • Matek405 not emitting some sounds?

    • Some fixes from Leonard

  • No plans for 3.6.8 so far

UTC1128 - Randy and Rover

  • Would like help getting patches back into Rover from master on some of the more important fixes

UTC1132 - Partner’s meeting tomorrow morning

UTC1134 - coverage

UTC1137 - issue report for 3.6 people running out of memory

UTC1138 - Septentrino next-gen GPS is awesome on reacquiring locks

UTC1139 - navigation

  • [11:39 AM] (Channel) akshath: Navigation libraries are common between Rover and Plane, Pierre Kancir suggested navigation controller selection to be a part of AP_Navigation library for Rover, but it would require changes with Plane code as well. What would u suggest?

  • Need to disable on 1MB boards, probably

  • Into AP_Navigation is OK: tridge

  • Auto-taxiing in Planes

    • Three people have asked tridge about this

    • Including someone who has permission to fly out of international airport

    • Needs auto-taxiing!

  • New waypoint type for taxiing?

UTC1143 - GSoc mentor volunteers needed!

  • New idea:

    • Waf build improvements

      • Speed builds up!
  • Fix Python3 problems!

  • 2 isn’t enough

UTC1147 - close