Dev Call June 6, 2022

Issues & Pull requests

Plane

Copter & Rover

GSOC

Attendee count (max): 19

UTC1101 - common.xml: add ignition_voltage to EFI_STATUS by peterbarker · Pull Request #273 · ArduPilot/mavlink · GitHub

  • Add ignition voltage to efi_status message
    • Merged

UTC1103 - common: add gimbal messages from upstream by rmackay9 · Pull Request #274 · ArduPilot/mavlink · GitHub

  • New gimbal messages
  • Are these messages good?
  • Randy thinks they’re not bad
  • Separate between device and gimbal manager
  • No gimbal manager messages brought in
  • Gimbal device messages only
  • Attitude status message comes in at 50Hz
  • Randy thinks these are an improvement in a few ways
    • Allows you to specify difference between earth and body frame
    • Pretty major missing feature
    • Point gimbal at target….
      • Then you want to lock onto that target
      • Then you turn vehicle
      • You want to lock the yaw in earth frame
  • Gremsy knows difference between Earth and body frame
    • Makes a difference to stability while making the turn
    • The gremsy takes care of pointing at the target
  • Current from-GCS messages don’t let you specify a frame
  • Messages are still WIP?
    • Shouldn’t be used in production….
    • 3 years….

UTC1112 - AP_Mount: custom Gremsy driver by rmackay9 · Pull Request #20915 · ArduPilot/ardupilot · GitHub

  • [9:15 AM] Peter Hall: If that is what the GImbal needs that is what we have to use, for now anyway.
  • We’re going to have to start removing features….
  • Peter’s worried about flash costs
  • This one needs guards to allow it to not be compiled in
  • Sensor suite….
    • Can we not have things compiled in by default?
      • Difficult due to sensor drift as things become available/unavailable
  • ToshibaCAN and KDECAN can both be compiled out on 1MB boards
  • [9:17 AM] Peter Hall: We could remove a bunch of QP frame types.
  • Angular limits come from device
    • Eases setup for user
  • New gimbal driver….
    • Doesn’t reuse existing storm driver
    • Gimbal device set attitude for the 4 different control regimes
      • Including “relax”
  • Manager vs driver?
    • Weird
  • This doesn’t make e.g. servo / storm handle the new messages
    • “Lock follow” should work with all backends into the future
  • Need to get James’ input on what he doesn’t like

UTC1127 - Copter, Plane, Sub, Blimp, AP_Mount: remove unused send_gimbal_report by rmackay9 · Pull Request #20899 · ArduPilot/ardupilot · GitHub

  • Removes gimbal report sending
  • We didn’t actually send stuff…
  • 100 bytes saved!

UTC1138 - INS: only run harmonic notch on active gyro by tridge · Pull Request #20107 · ArduPilot/ardupilot · GitHub

  • Don’t filter all IMUs at the same time
    • Reduce CPU usage considerably
      • Factor of 3 on CubeOrange
    • Most dev was done on single-IMU systems….
  • When not running filter reset is called all the time
    • When you’re chosen your filter starts on the first gyro sample
    • Noise will get through for some number of milliseconds after switch
      • Should be small as still running low-pass filters
    • Do attitude control reset?
    • Notch filters are fairly short time-constant….
    • Post-filter logging might require them all to be run
      • There’s an option for that
      • Need to document that option
  • 9:41 AM] Peter Hall: we could do a attitude control reset
  • [9:41 AM] Peter Hall: like we do for pos control with EKF
  • [9:43 AM] Peter Hall: A jump in rate will give huge D term.
  • [9:43 AM] Peter Hall: I guess we have D term filter.
  • Needs flight testing
    • With RC switch to switch IMUs….
    • Probably on a Copter
  • EKF used to use the filtered data for on-ground/is-flying
    • Now can’t even get the filtered data
    • Only consumed by rate controllers
      • Only ever use one gyro
        • Could potentially do median filtering on 3 gyros
          • Might be smoother…
  • We need to claw back ram and CPU on CubeBlack
    • Octaquad with per motor notch filtering….
      • 72 notches…
      • Correct way to get good low noise
        • Throttle-based filter gets a spread of freqs if you have any rpy imbalance
          • Including from wind
          • Notch filters don’t do well here
          • Fft doesn’t work because of spread of freqs
          • Need per-motor rpm-based
  • Randy will test

UTC1147 - Plane: Quadplane: QRTL climb to QRTL alt first if in Q mode by IamPete1 · Pull Request #20645 · ArduPilot/ardupilot · GitHub

  • Tested by Alex and Henry
  • QRTL climb to QRTL alt first if in Q mode
  • Henry will test some more…
  • Qpostcontrol state reset might be a good idea for a follow-up PR

UTC1152 - AP_Periph : add AP_Stats for Node statistics monitoring by Pradeep-Carbonix · Pull Request #20858 · ArduPilot/ardupilot · GitHub

  • AP_Stats in AP_Periph
  • Merged!

UTC1157 - AP_WheelEncoder: fix WheelEncoder_Quadrature timestamp. by MagicHardware · Pull Request #20885 · ArduPilot/ardupilot · GitHub

  • Quadrature timestamp fix
  • Us vs ms
  • Might affect tuning on balancebot…
  • Massive gains might need reducing
  • Maybe backport?
    • Perhaps not with the tuning problem

UTC1159 - Plane: Quadplane: force conversions of Q_M_PMW_* params if invalid by IamPete1 · Pull Request #20892 · ArduPilot/ardupilot · GitHub

  • Plane: Quadplane: force conversions of Q_M_PMW_* params if invalid
  • Potential backport
  • People are still having troubles…

UTC0003 - AP_Motors: Resolve Issue 20894 by muramura · Pull Request #20900 · ArduPilot/ardupilot · GitHub

  • Merged!

UTC0004 - CAN: Support DroneCAN EFI periph node and driver by tridge · Pull Request #20901 · ArduPilot/ardupilot · GitHub

  • Uart to CAN
  • CAN: Support DroneCAN EFI periph node and driver
  • Very, very redundant aircraft are now possible…
    • Redundant motor layers, redundant EFIs, redundant ICE, ….
  • nread thing?
  • Peter will review after call

UTC0019 - Support high numbered ESCs in ESC telemetry by tridge · Pull Request #20918 · ArduPilot/ardupilot · GitHub

  • More esc telemetry messages
  • Fixes rotation issue getting mavlink messages out
  • External tools one ESCs 1 to 8?!
    • We rejected Tom’s PR to do the shifting
    • Odd that we burn autopilot bytes wit the remapping
    • We’ve had multiple requests to do this by end users
  • Parameters for esc telem
    • ESC_TLM_
    • ESC_T_?
  • Cp command has changed under Windows, breaking cygwin
    • If you cp arduplane to arduplane.elf it used to add a .exe helpfully
      • Upgrade, Windows no longer does that

UTC0033 - Tools: cope with change in cygwin build by tridge · Pull Request #20926 · ArduPilot/ardupilot · GitHub

  • Fix cygwin build
  • Merged!

UTC0037 - Plane update

  • Almost time for 4.2.2
  • Many pending
    • Lots of bugfixes
    • Probably dual harmonic notch will go in
      • Jaaky raised some questions around it
      • Arming-require-zero is an issue but already in 4.2.1
    • Possible beta this week
    • Replay fix to go in
    • Probably not the ringbuffer implementation just yet
      • High-rate GPS issues?
        • What happens with the two different buffer implementations
    • Q_navalt_min might come in
    • Dynamically allocated pool size?
      • CubeBlack/PH1 and CAN?
        • Memory problems
        • Terrain wasn’t initialising
        • Arming failure on Copter
        • CAN memory use changed
          • malloc/free used to be used on every packet
          • Now we pre-allocate a packet
            • Because failure to allocate at runtime would be bad
            • Using pool allocator
            • 8kB per bus on Black
            • This is in master
              • Should we bring it back into stable?
              • New memory allocator…
                • Some better internal protection
          • Probably need a Wiki page on how to save RAM
            • Symptoms and fixes
              • E.g. terrain, ftp, logging
              • This would add another knob
                • Can actually go very, very low on these numbers
      • Should it go into stable?
        • Stable Aug 5?
        • Worth deferring?

UTC0037 - Copter +Rover update

  • 4.2.1 stable is on its way out
  • 4.2.2 will follow along with Plane
  • No pressing issues
    • Andy’s got a few like d-shot LEDs
    • The crsf issue test out the new crash dump stuff which worked very, very well
    • PR to save the crash dump to SD card on boot
    • Parsing bug / off-by-one bug
      • Expresslrs

UTC0055 - GSoC update

  • Still in bonding period
  • Two students have already started
  • Two students still need to get engaged
  • Rishabh and PH’s students….
  • Hardware requests inbound

UTC0056 - nokov optical tracking issue

UTC0003 - predicted motor failure message from lua

  • Worked well…
    • Warning went away after a while?
    • Would be nice to set the motor fail bit from LUA
    • But repeating the statustext would be much easier
  • Having lua script stomp over the state of the vehicle and not know where stuff comes from is a little interesting…
    • C++ or LUA?

UTC0103 - close