Dev Call Aug 20, 2018 2300 UTC

**Issues & Pull Requests


  • Final Update

Open Drone ID


  • 3.9.1beta2 Update


  • Update


  • Update


1 Like

Attendee count (max): 19

UTC2301 - Recursive semaphores

  • Semaphores that a thread can take multiple times but must release the same number of times
  • Stops you deadlocking against yourself
  • Adds against all the HALs
  • Please check comment on PR discussing the locked-interface pattern
    • Caller grabs the lock on the library
      • They can call whatever they want on the library at that point
      • When scope closes lock on library is released
      • Caller takes care of atomicity
      • Updates on library are happening on main thread but only update routines need to take lock
  • Tridge is currently using this for new avoidance code
  • MdB: I’m not sleeping, but I’ve talked with tridge a bunch before this :smiley:
  • [9:10 AM] (Channel) MdB: He cost me my lunch before the call actually :smiley:
  • ML: I have no idea what any of this means, so I think it’s a great idea!
  • Maybe wait for FF or Lucas to comment

UTC2313 - Stack overflow checking in SITL appears to have broken MacOS

  • Probably memalign vs posix_memalign

UTC2314 - avoiding bad RC trims

  • Trimming planes in manual with RC trims means your FBWA may not fly straight (commanded)
  • Add arming check that RC is in deadzone for 2 seconds since boot
  • Should we check all channels rather than just the primary flight channels?
    • Could be painful
    • But would ensure correct model is selected
    • Be at switch
    • MdB: IE my PX4IO switch moves with flight mode switch
    • [9:19 AM] (Channel) MdB: which means it may not be trim
    • Maybe sometime later
  • Sounds good, just needs implementing!

UTC2319 -

  • Unit test suite should set the pragma indicating direct equivalence testing is OK
  • Digression into hatred of templates
    • Stop using templates so much?
    • Is_equal a template for floats and doubles?
  • Wrap180 should not be a template
    • Scaling factor
    • Critical function called rapidly in loops
    • Or make a specific float version

UTC2324 -

  • MdB likes it
  • Tridge likes it
  • Can be merged

UTC2330 -

  • Tridge likes it
  • MdB likes it
  • Can be merged
  • Catches where you use the wrong ID in a block going to disk
    • Most of the time

UTC2333 -

  • Magic constant question
  • Upper and lower limits?
    • Lower limit 1 parameter
    • Some other upper limit
  • Is worth pursuing
  • Solo has 95% lost packets
    • Packet injection issue due to consecutive sequence numbers?
  • Scale all your like with flow control
    • Not above 115kbaud otherwise

UTC2343 -

  • Tested in autotest
  • Randy thinks it looks good and can be merged

UTC2345 -

  • Servorelayevents
  • Only on unused channels
  • Perhaps a specific “relay” or servo function in addition to None
  • Mavlink control?
  • Put it in as-is
    • Can add another define later if we feel like it

UTC2349 -

  • Small optimisation
  • This function should be using .get()
  • Pre-existing bug discovered live!
    • Assigned to Michael

UTC2355 - 8690

UTC2358 -

  • Yaw from GPS
  • still pending, won’t be done in a little while due to the OBC
  • Removing devcalltopic

UTC0000 -

  • More accurate timestamp
  • Could end up logging too early
    • Potential race condition
    • Not easy to fix this race
    • Maybe change semaphore to blocking-take
      • This would not be safe as it’s in interrupt concept
        • Aircraft would fall from sky
        • Tridge will add a check that you’re not blocking-taking
  • Not in use
  • Pending for far too long
  • Tridge doesn’t immediately recognise the code
  • GPS time-lag estimation PR?
    • [10:06 AM] (Channel) MdB: the main debate on it was what to do with units that are over-configured?
    • [10:06 AM] (Channel) MdB: but I maybe can’t recall…
    • 10Hz but can’t do 10Hz needs to reconfigure the GPS back to 5Hz
  • [later] tridge worked on the patch, asks Nate to test

UTC0011 -

  • Leonard’s started to look at it
  • Will be a parameter under MOT_SLEWRATE or similar

UTC0014 - GSoC

  • Coming to an end
  • Students did very well
  • Jaime is working through Sephyr’s work
  • New repos on the ArduPilot github list
  • All evaluations are in already
    • All passed except the student who had to withdraw for legal reasons
  • Thanks to all mentors and students!
  • Mentors meeting in September
    • Right before the China conference

UTC0018 - open-drone-id

  • Intel is on workgroup with FAA
  • Bluetooth beacon packet with 26 or so bytes
  • Similar to ssid on wifi, always being sent
  • 400m with bluetooth 4.2
  • Should be able to do a km with bluetooth 5
  • Two messages at two different rates
    • More-info
    • Position
  • Tom’s working on mavlink gateway
  • All in the blind
    • No connections
    • Alt, vel, pos
    • 32-bit number
  • Working group is authoring it
    • Intel is kind of the secretary of the working group
  • [10:34 AM] (Channel) TP: yes, there’s a speed vector
  • TP: and AGL and MSL and an advertised accuracy
  • TP: Open Drone ID pics:

UTC0030 - plane update

  • New beta out
  • Change of default pid gains for roll/pitch
  • Old values way too low
    • Raised them
  • Avoid Soggy planes
    • Want first throw-off experience to be better
  • Bad experience on first flight when plane won’t pull up!
  • TECS_PITCH_MAX of 15
    • So automatic navigation controller doesn’t have bad landings due to poor pitch landings
    • “Balloons” up and loses too much airspeed
  • L1 rate changed to tighten navigation up
    • 17
  • MdB: oO I thought it was supposed to be ~1.5x speed?
  • Autotune for TECS?
    • Not yet
  • OSD updates in 3.9.1
  • Force messagewriters fix for very low dataflash buffer sizes

UTC0038 - Copter/Rover update

  • 3.6 testing
  • Randy was away last week
  • Another rc coming this week
    • Rc8
  • Fix for allowing do commands before takeoff command
  • Boost working or not?
    • Not initialised correctly
  • Growing calls for a Rover release
    • Omniboats
    • ChibiOS-only boards etc
  • Sailboat code is moving along nicely
    • With iampete
    • Testrun in next week or so
  • Non-gps navigation is moving along slowly
  • When using external positions sources you can’t really use your compass
    • Need to turn them off
    • Otherwise your frames get misaligned
      • EKF will align with compass
      • External system won’t align with that

UTC0041 - tiziano

  • Willing to make some media videos to make on our ArduPilot page if we can get him to the interdrone conference
    • TP: Made some great videos in the past
  • TF: I should have joined the dev call much sooner! Thanks for making my life easier!
  • A lot of work to put a nice video together
  • Funding request!
    • How much, benefits etc etc
    • Will do!

UTC0050 - Believer planes

  • Still not coming
  • Olivier still poking them