Dev Call June 10 2019 2300 UTC

**Issues & Pull Requests"





Attendee count (max): 14

UTC2301 -

  • Remove unit from parameter name (_CM)
  • SI vs non-SI units
  • We really shouldn’t use cm
  • Not all GCS show unit next to parameter
  • Cause more heartache than it will save
  • No compelling reason to make this change?
  • New units should be in SI units
    • Old library and a new parameter? Do you use old or new?
  • If you are forced to do a unit in non-SI then tacking the unit on means when doing a conversion you can use the correct name name later
  • OP has a nice Wine installation of Mission Planner

UTC2311 - Azure

  • Either Francisco or tridge needs to look at this one
    • Python is missing!

UTC2311 - add -Woverflow

  • Merged!

UTC2313 -

  • EKF reconcile lane switching

  • Vehicle popped parachute when not needed

  • One lane was unhappy but other was happy

  • Would be nice to reconcile the two algorithms

  • Copter requires 10 samples the same in a row for EKF failsafe

  • Add hook for EKF to consider lane-switching

    • But not if changed in the last 5 seconds
  • Code is adjacent for the two different lane-switching mechanisms

  • EKF sensor affinity code is related

    • But will be opt-in as you need good sensors
  • Paul to review this before this is merged

    • Tridge will ping him

UTC2320 -

  • Merged

UTC2326 -

  • On IO processor
  • Crosstalk?
  • Going into failsafe
  • Author thinks failsafe may have been active
  • If we do go into failsafe why do the motors start?
  • Oneshot on IO processor is a thing
  • Try increasing timeout and see if that makes a difference?
  • This is 3.6.7
    • NuttX?
    • The IO firmware on 3.6 is still the NuttX one

UTC2332 -

  • Should we use a flag on parameter instead?
  • 16 bit worth of flags
  • Only using 6

UTC2336 -

  • Can be merged after we change “and” to “&&”

UTC2336 -

  • May prevent people from arming when they get an FPE
  • Checking result vs inputs
  • Is_positive vs is_negative all over the place
    • We don’t check for isnan and inf etc etc
    • Exceptions?!
  • Coordinate with isnan and isinf?
  • Find construct for doing good programming practice?
    • Some sort of check vs fpe caught check?
  • No determination on this one

UTC2358 -

  • Incomplete so far

UTC2359 -

  • Several bugs unearthed
  • Probing angles wasn’t quite right
  • Whole bunch of AP_Math changes weren’t actually needed for merged code
  • Why moving implementations from .h into .cpp?
  • Not waste flash space
  • Faster compilation
  • Complex functions in .h not good
    • Functions calls are relatively cheap on stm32
  • Even just a single multiply and it should probably go in .cpp
  • Conversion back and forwards between lat/lon and Vector2f?
    • Particularly proximity
  • Dijkstra memory allocation still an issue
  • Randy will merge this within next few days

UTC0007 -

  • Merged!
  • Copter namespacing fixed

UTC0009 -

  • Updating default parameters…
  • Tridge is discussing with people on why he thinks this is a bad idea
  • Tridge is concerned about vehicles falling from the sky with this patch

UTC0010 -

  • Have to make sure Parameters and log download speeds are not affected
  • If there’s no impact on speed then this could be merged

UTC0017 -

  • Will move SurfaceTracking into the Copter object rather than into the mode object

UTC0023 -

  • No Pierre
  • Need a comment indicating whether it should be merged or not from DevCall tagger if they’re not on the call

UTC0024 -

  • Needs a rebase
  • Commit message fix
  • Commented

UTC0025 -

  • Merged

UTC0027 -

  • Seems reasonable to tridge

UTC0028 -

  • Had a bad flight on this PR
  • Can’t afford to lose Paul’s work
  • This PR was originally done for SkyRocket
    • SkyViper flight
  • Needs testing on multiple flights on a Copter
  • Need a few flights on an Iris
    • Randy will take this on

UTC0033 -

  • Do we want this in Copter?
  • Adds DO_LAND_START for Copter
  • Complementary solutions to smart RTL
  • Randy is happy for this to go in
  • Peter to rebase and write a test for

UTC0037 -

  • Handling of compass should be improved
  • If you see oddities let tridge know
  • Constrains learnt magnetic field vs earth magnetic model
  • Hard to estimate z
  • Compass affects attitude estimates
  • So this should improve compass a lot
  • Major change in the way that we handle the compass

UTC0037 -

  • Power control on CubeBlack
  • Unresolved issue with other PixHawk1 boards
    • We think
  • Now change should only be CubeBlack
  • Really need to get tridge a failing board
  • Stuck in bootloader means not running this code
  • Could notionally be a MissionPlanner bug
    • Sending boot ahead of time would have this effect
  • Video was good
    • But a screen capture or script output would be good

UTC0047 - GSoC update

  • Peter Hall and sailboats
    • [10:36 AM] (Channel) RM:
    • [10:36 AM] (Channel) RM: peterHall’s sailboat with motor ^^
    • Windvane working well
      • That’s the thing on top of the mast
      • Ultrasonic anemometer
    • Few issues to work through on motoring
    • Want to estimate lift/drag soon
  • Askhath’s got a MAVProxy PR merged
  • Tridge will be working with Rajat on AirSim stuff this-afternoon
  • Great progress with vio stuff

UTC0048 - Plane update

  • Nothing specific
  • Maybe missing a reset of integrators on rate controllers

UTC0050 - Copter/Rover update

  • Should start 3.6.10 beta soonish
  • 3.7 beta testing 6 weeks away yet
    • Stay-out zones really wanted
      • Tonnes of features/improvements otherwise

UTC0052 - new firmware server

  • Old one slowed down by a factor of 3 for no apparent reason
    • Consistent slowdown
  • New server
    • CPU-optimised
      • Tied CPUs
    • New version of Ubuntu
  • Actually completing builds now
  • Build_binaries also got patches to make it faster

UTC0055 - Comparing a uBlox Zed-F9P GPS with a M8P without external corrections

  • GPS comparison
  • Ublox F9 will make a huge difference in GPS quality
    • Primarily dual-band stuff
      • Phase information was used to use L1/L2
      • new L2C is unencrypted and can be used for dual band
      • Two bands on two frequencies can be used to estimate the ionospheric disturbances and correct for them
  • More testing for RTK coming
    • Maybe more resilience to multi-pathing
  • Pretty much everyone will be moving to these in short order
  • Others have done much more detailed analysis
    • RTK fixes are better
      • Tridge is planning on reproducing today with nastier setups
        • RTK does cm-level accuracy with clear sky views
        • Pretty bad if you don’t have those conditions
  • Jeff’s done comparisons
    • Clouds roll in on single-GPS your lock goes to hell for quite some time
    • L1/L2 lock gets much smaller much more quickly
      • Very good within 30 seconds
    • Losing RTK while flying - how bad?
  • Antennas are heavier with L1/L2
    • Maxtenna etc are working on tri-band antennas (L1/L2/E5)
      • [11:04 AM] (Channel) JW: the F9 is E5b specifically
    • 25% heavier
      • US$290
      • In sort of range typical user can afford
      • More should come to market soon and perhaps reduce price further
      • Vs US$2400+ L1/L2/E5/Helical delivered in Sep 2018
      • Would like to compare Septentrino vs this
        • Septentrino are going to compete
      • Tridge will get one of these soon
        • His current one was impractical for flight
  • Probably need to rejig some EKF thresholds
    • GPS trust thresholds
    • F9 accuracy numbers are more believable
    • M8 were fantasy
  • Outperforms M8P by a lot even with single-band antenna
    • Better silicon?
    • Better receive sensitivity?
    • More channel tracking?

UTC0110 - thanks to B. Olivieri

  • For lots and lots of great work on the wiki

UTC0110 - Close