Dev Call March 23, 2020

Development Team
2

Attendee count (max): 25

UTC1100 - Zoom and the conference

  • Everyone needs to disable their video
  • Except the presenter, who should enable it

UTC1108 - https://github.com/ArduPilot/ardupilot/issues/13868

  • Bad buzzer tone string causes reboots
  • Seem to have found the problem already
  • Can be reproduced in SITL
  • Somebody’s going to look at it… MdB PH or PB

UTC1111 - https://github.com/ArduPilot/ardupilot/pull/13791

  • Add support for do-cam-trigg-distance
  • If you set third packet to true then the vehicle should take picture immediately rather than after travelling preset distance
  • Why not take-photo directly?
  • Skip extra variable

UTC1120 - https://github.com/ArduPilot/ardupilot/pull/13870

  • Azure fix VMImage
  • Merged…

UTC1120 - add sanity checking for values in parameter documentaiton

  • Merged!

UTC1127 - https://github.com/ArduPilot/ardupilot/pull/13820

  • Increasing size of arrays in Dijkstra wasn’t working
  • ChibiOS was missing realloc
  • CAN GPS was causing issues?
  • More bugs since
  • Only doing realloc between main heap
    • Vs second eap
  • Built-in realloc would not have zero extension.
  • Current implementation maximises fragmentation
    • ChibiOS allocation system for lua SITL
  • This fixes a real bug but will fragment memory
  • Realloc_type or std::realloc?
  • Can rename realloc later
  • Merged!
  • Needs to be backported

UTC1132 - scripting - support sending/receiving mavlink messages

  • https://github.com/ArduPilot/ardupilot/pull/13660
  • Queuing of outgoing messages?
    • May currently starve other senders
  • Queue in size of bytes instead of messages?
  • Streamrate based on ap_message_id rather than mavlink id?
  • Need an example script….
    • E.g. named-value-float
  • Need documentation in Wiki for how to do the mavgen
  • Limit bandwidth in kB/second

UTC1159 - https://github.com/ArduPilot/ardupilot/pull/13835

  • Add logging of rc protocol bytes

UTC0011 - https://github.com/ArduPilot/ardupilot/pull/13793

  • Virtual FS usable via mavlink
  • Ftp get @PARAM.pck
    • ~12kB and is full parameters
  • @MISSION/mission.pck
  • No ability to set parameters yet
  • Parameter stuff is coming later
  • Randy to test

UTC0020 - Plane update

  • flaperon fix for Plane
  • Still planning on syncing with Copter for next release

UTC0021 - Plane 4.0 release

  • If you want to test your gains you have to test them in autotune mode
    • Have to switch out of autotune mode and back in and then you’re using the new gains
    • What behaviour should we have?
      • Switch to use autotune gains?
      • A switch for banks of parameters?
      • SkyViper 2018 had banks
        • Three different sets tame, sport, extreme
        • Dozen parameters
        • Three banks, three copies
      • Create trees of parameters in scripting?
        • Atomic way to store parameters?
    • Logging issues with losing messages in dataflash-over-mavlink
      • Probably running out of bandwidth
  • Fast sampling causing overruns
    • Underlying issue
      • Table of function and expected time
        • Expected time doesn’t account for things like SPI processing
        • Transfers and filtering in background in high-rate IMU stuff causes the main thread to stop while it is processing those
          • Causes overruns as we decide we can fit stuff in but then we overrun because the interrupts take time
        • Boost main thread priority above SPI thread until we get to where we run the EKF
          • But we start getting overruns because we’re not reading SPI fast enough. We get resets and whatnot. And the FIFO corrupts as the sensors are buggy
    • Couple of small tweaks
    • Defer the filtering?
      • More memory
      • Invasive
      • Something to look at post-conference
    • Move EKF into a different thread?
    • Mavlink into a different thread…
    • Invensense config bit stop filling FIFO?
      • Doesn’t work?!
      • Corruption regardless
      • We do say stop….
      • No reflection in depth of FIFO
        • We use an experimental guess at the length
      • Maybe if you’re only taking one sensor it works?
    • Welcome to Lucas!
      • Working from home!
  • CubeOrange and mRo boards and passthrough
    • Buffers stuff things up
    • GPIO speeds?
    • Capacitance level on pin affecting rise time?
    • Analogue filtering on the pins?
    • PK discovered if you take away the hardware buffers things worked perfectly
    • Check filter options?
    • Check GPIO options?
    • Convenient function to save/restore state of pins

UTC0044 - Rover

  • No much
  • But new APSync stuff coming
  • Repeated requests to sort out navigation on Rover and Copter
    • Scurve stuff!

UTC0044 - conference

  • Up to 100
  • Would be bad if speakers could not join…
  • May have to turf people out….
  • Or run as a webinar?
    • Or buy 10 hosts?!
    • $300
  • Stream it live on youtube?
    • Obs

UTC0050 - close

1 Like