Dev Call Feb 4 2019 2300 UTC

Issues & Pull Requests





Attendee count (max): 19


  • Run 3 EKFs of same type?

    • Logging is a problem

    • But should work OK past that

  • Using extra CPU

    • How do we use extra CPU we’re now getting on e.g. H7s

      • Haven’t we got this on Linux already?

        • Linux doesn’t offer high loop rates

          • E.g. SPI rates

            • Call overhead
    • Should we bump Copter loop rate up from 400?

      • Maybe 750?

        • Stay below 760 for crappy lsm
      • Can’t run faster than your slowest gyro

      • May be stability advantage on small Copters

      • Easier tuning, apparently, from betacopter reports

        • Tridge needs to talk to Leonard

          • True for our arch?
    • Neural network?

UTC1107 -

  • Randy to look at

UTC1115 -

  • Merged

UTC1119 -

UTC1121 -

  • Do we merge this?

  • Prove no harm?

  • Thinking and working required before this is merged

  • Do we merge it and fix things later?


  • Some code formatting stuff

  • Not completely obvious it does nothing when not enabled

  • Tridge will take this, but it will almost certainly bitrot

UTC1125 - fix support for external nav


    • Adds support for input of external positioning
  • AL: why so many changes

  • RM: found some issues with integration of external navigation integration

  • Rebased recently

  • This has been retested

    • Found some problems

      • Fine for a while then loses the plot
  • Gives position and quaternion attitude estimate

    • Fusing (or not!) of the attitude estimate may be the problem

    • Could be “local” compass heading issue

  • Fail in simulation?

  • Could use one of the ekfs to loop back into another to test

    • So ekf3 uses ekf2 as an external source

    • Logs would give you a good way to validate it

    • Would need to add lag

  • Will run extended test in SITL to see if it replicates

UTC1133 - funding for EKF development?

  • Randy has been trying to get to some PRs

    • But feels underqualified
  • We fund a bugmaster

    • Perhaps we should fund some ekf dev work?
  • Will talk about it at dev conference

UTC1135 - AP_HAL_ChibiOS


  • Timer issues

  • Drama in gitter channel

  • Pin mappings for aio different to the non-aio board

    • New hwdef?
  • Happens in other boards too

  • Seen Betaflight motor-ordering PR?

    • Betaflight vs dji vs ardupilot

    • We chose the pinouts so quadx with default motor ordering would work out of the box

      • But when your pins are in the corners of your boards next to the motors, it’s important to use those pins for that motor

      • Should follow silk-screen ordering

  • Incorrect silk-screens are out there.

  • Pinouts change between different versions

  • Maybe hwdef.dat should allow setting of frame-type-default

    • E.g. set to betaflight-x on this hwdef

    • Already some people flying these boards

      • With patch as-is we’d have people crash

        • If we use the frame-type defaulting then things should still work

          • Only works for X

            • Not v or h
  • Will need to warn Plane users on upgrade

    • Modern ESCs should not initialise, but we need to warn
  • Currently our servo outputs don’t match the silkscreen

    • We are going to match the silkscreen where possible

    • If two boards with different silkscreens then two different hwdefs

  • Need to get three patches in:

    • Kaktute f4/f7 patch, this motor-ordering PR and a patch for #ifndef HAL_DEFAULT_FRAME_TYPE etc

    • Tom and Josh will look at this

    • Need to be careful to avoid changing the timers

      • Pin number and timer number don’t change

      • Change things in parentheses


      • defaulting-defines on config.h

      • Hwdef can then override the frame type

      • Set frame-type to 12 for kakute (blheli ordering)

  • [10:51 AM] (Channel) RandyMackay: it depends on the drone model/controller.

If we’re speaking RTF models like the M600 / M210, these are special. On the S900/S1000 motor order is a non issue when using ardupilot

  • RandyMackay: When asked why it was a “non-issue” for DJI, Luis replied, the motor controls are located on a small connection board, so it’s just a very simple matter of swapping a few jst plugs

UTC1152 -

  • Tridge, Peter and Jacob to look at it

  • Badges would be good.

    • Anybody done it?

    • Jacob will talk to William

UTC0004 - plate update throttle mix

  • Nice piece of work

  • One transition case that needed looking at….

  • Will put this in and mark it as needing backport

UTC0006 -

  • Merged!

UTC0008 -

  • Once MdB and Peter are happy it can go in

  • Peter explained idea about “dirtying” messagewriters so they run again

UTC0011 -

  • Tridge isn’t quite ready to review this yet

UTC0012 -

  • Configurable waypoint speed

  • stale?

  • Devcall removed

UTC0013 - partner’s call moved to next week

UTC0013 - partner’s call moved to next week

UTC0016 - Onion update

  • Never going to go in if tridge doesn’t test it

  • Waiting on tridge

UTC0018 - learn throttle trim

  • Lots of implications within TECS

UTC0020 - Randy and copter update

  • Copter-3.6 went out for beta testing

    • RSSI indicator for ChibiOS boards in 3.6

    • Parameter description changes

    • Some bugfixes

    • Skipping-first-spline-waypoint fixed

    • Dt calculation was incorrect 2/3ms issue

    • Rc override fix

  • Some 3.6 issues remaining


    • Matek 405 logging not working

      • Working for iampete
    • Compass not working on some board?

      • Something to to with rc parsing?!
    • Sumd protocol not working?

      • Graupner

      • Never worked before

      • We do claim it works now…

UTC0024 - Rover update

  • 3.5.0rc2 went out for beta testing

  • Rover-3.5.0-rc2/-rc3 is available for beta testing

  • Nobody has complained about anything

  • One bug Jaime found

    • Tiny turns would give a divide-by-zero

    • Is_zero is not always sufficient, it turns out!

      • Perhaps it was just misplaced, as the variable was munged before being used as a denominator
  • Probably no -rc4

  • More documentation on Wiki on ROS stuff


      • Little 3D viewing tool for ROS when it is building up a map
    • Next step is to add some navigation

      • Jaime’s been doing some work on that

      • Offload high-level navigation to ROS?

UTC0028 -

  • RPM implementation for Rover

UTC0032 - GSoC update with Jaime

  • Project Application deadline Feb 6

    • Will get application in soon

      • Same mentors as last time and adjust later

    • Look at it!
  • Emphasis on iterating existing code

    • So our devs know what they’re talking about!

    • More likely to make it into master


    • March 25th starts application period

    • April 9 to May 6 is application review period

    • May 6 student projects are announced

    • Dev period starts May 27 to Aug19th

  • Randy will nudge our website a little