Dev Call May 27 2019 2300 UTC

Craig not able to attend or post an agenda. Randy stepped in to chair. Thanks Randy!
Craig will be back next week.

Attendee count (max): 15

UTC2301 - https://github.com/ArduPilot/ardupilot/pull/11416

  • Merged!

UTC2313 - https://github.com/ArduPilot/ardupilot/pull/11403/

  • Redo as a single function with an if statement in the top

UTC2316 - parameter documentation

  • Should we have SITL in the list or not?
  • Merged!

UTC2317 - https://github.com/ArduPilot/ardupilot/pull/11401

UTC2320 - https://github.com/ArduPilot/ardupilot/pull/11390

  • New file should get block comment for what the file is for
  • Which way should we move mavlink parsing things?
    • Into libraries
    • Or out into mavlink
    • Trend is for libraries to do their own stuff
    • Permanent dependency you can’t get rid of
    • GCS_MAVLink primarily infrastructure to talk to ground station
      • Break out a separate file in AP_Mission instead?
    • Half of AP_Mission is mavlink conversion code
      • Which annoys MdB
      • AP_Mission_MAVLink.cpp
      • Harder to refactor AP_Mission to be internally correct?
    • Working in GCS_MAVLink
    • Amilcar like the way this is moving
  • Killed this PR, will move code in the other direction

UTC2330 - mini-onion WindVane

  • https://github.com/ArduPilot/ardupilot/pull/11344
  • Tested in SITL
  • SITL backend! Yay!
  • Calibrated by twisting the windvane
  • Calibration is once, not once per voyage
    • Device isn’t that accurate anyway
    • min/max are not rescaled in use
    • Worry about voltage spikes?
    • Ratiometric?
      • Power supplies vary
      • Sensors that are ratiometric assume a base voltage so you have to scale according to input power
      • Regulated or not regulated?

UTC2336 - quadplane precision v2 by Luke Dempsey

  • https://github.com/ArduPilot/ardupilot/pull/11339
  • K_param_precland 256 may not work well
  • 512 top-level parameters was added a while ago
    • Not backwards compatible
    • Above 255 will work on new code but not if people go back to really old versions of the code
      • Should we open up the higher range?
      • Probably safe after 3 years?
    • Only a problem if you go back from new to old code
  • Needs a detailed review
    • And a rebase
    • Tridge will try to find time to review it
  • Not pre-onion?
  • Tridge will rebase it
  • Perfectly supportive
    • Just needs time to look at it

UTC2340 - https://github.com/ArduPilot/ardupilot/pull/11339

  • I2c transaction count and spi transaction counts via mavlink?
  • Merged!

UTC2345 - io processor resets

  • Tridge also saw something similar to JC
    • But he was flashing
    • Maybe checking version of firmware at the time?
      • Check timestamp of io processor should allow for io processor to have been up for some period of time
      • Need to detect IO processor reset so we send across all of the configuration
    • It should never need to reset so it should currently be an internal error

UTC2348 - tradheli

UTC2349 - battery failsafe as percentage?

UTC2350 - Plane and tridge

  • Proposal for airspeed-only-if-no-GPS?
    • Needs thought
  • Sensor affinity could be nice for both airspeed and GPS
    • Tridge is working on
    • [9:52 AM] (Channel) MdB: Because that scenario encourages not calibrating/validating the sensor then you swap to it in the failure mode
    • [9:52 AM] To Weekly devcall: @MdB calibrate in-flight from estimated wind speed :stuck_out_tongue:

UTC2351 - Copter and Randy

  • 3.6.9 released
  • 3.6.10rc1 including RC logging and watchdog changes coming RSN
    • Estop switch issues?

UTC2353 - Rover and Randy

  • https://www.youtube.com/watch?v=bS5orrYR4C0
  • Bendy ruler algorithm from CUAV
  • Patches coming in RSN
  • Dijkstra’s too
    • “Safe nodes” generated from polygon fence
    • Inner fence x metres inside the polygon fence
    • Each point on polygon fence gets an inner fence
    • Visibility graph
    • Runs in a separate thread
    • Runs pretty well
      • RAM and CPU usage should be checked
  • Tridge has been considering adding more bends into the ruler
    • So it can take more turns before getting somewhere
      • This many turns not an issue on CUAV
  • Dijkstra’s may collapse when you have hundreds of fast-moving objects
  • How do you do moving objects in Dijkstra’s?
  • Identify safe places how?
    • Safe is dependant on time
      • Use projection and check for lines crossing to determine safe places
    • Turning east vs west may get you somewhere at a different time
  • Dijkstra’s currently runs once per destination
    • Stayout zone could be a Plane’s projection
  • Moving spheres in 3D space
  • Bendy ruler is a bit like a 2-ply-search in chess
    • Game tree
    • Chess-search algorithms might be applicable
      • Might be able to use relatively small amounts of CPU to get 8 or 9-ply searches
  • No projection based on velocity and turn rate hasn’t been included
  • This is built into AC_Avoidance not AP_Avoidance
  • Things are going into separate classes
  • Iampete is hoping to look at this as part of his GSoC too

UTC0005 - sailboat

UTC0007 - close