Dev Call June 3 2019 2300 UTC

Issues & Pull Requests





Attendee count (max): 15

UTC2301 -

  • Merged

UTC2305 -

  • How does this interact with threading?
  • How about F7 and H7
  • Maybe part of an example?
  • Stop clearing the bits before reading

UTC2313 -

  • Allow do-set-servo on rc passthrough
  • Control servo output with both pass-through and with do-set-servo
  • Ordinarily the pass-through overwrites the do-set-servo
  • Adds a little deadzone so that if the RC passthrough isn’t moving then the do-set-servo remains
    • Will work kind of like mode switches
  • Passthrough works for channel a to channel b
  • Some integer underflow issues on line 51
  • Happy with the concept
  • May need a bitmask for the channels we’re locking within the the deadzone

UTC2321 -

  • Current structure because we didn’t have singletons
  • Density of the air
  • True air speed is analogous to groundspeed and used for navigation
  • Airspeed is airspeed…
  • Should getEAS2TAS be in AP_Baro?
  • Get rid of calls on theon the airspeed sensor and replace with call on the AHRS
  • davidbuzz: do any gps modules have a baro onboard?
    • Yep (zubax)

UTC2329 -

  • Merging ArduCopter.cpp and Copter.cpp
  • Historical reason we still have ArduCopter.cpp
  • Should we do this for all the vehicles

UTC2330 -

  • Real bug when switching lanes
  • Copter and quadplane in VTOL mode fly with positions relative to EKF origin
    • InternalNav is relative to origin
    • We assume there’s one EKF origin
    • There’s actually one per lane
    • Usually they’re close together
    • They can diverge if you get GPS glitches
    • If you do a set-origin via mavlink it will only set the first one
      • Usually you don’t set-origin if you have a GPS
      • Hopefully you don’t
  • Put an origin in the top-level NavEKF2 and NavEKF3
    • But that will be bad if we fall back from 2 to 3
  • Put it into the AHRS instead
    • DCM has home but not origin
  • If EK3 gets origin first but you’re actually using EK2?
    • Only set at AHRS level if you are the active estimator
  • PR coming
  • The EKF drift mechanism
    • For long flights with baro drift
    • Assumes the GPS is long-term stable
    • It logs NKF?.OH in the logging which logs the origin height
      • It wasn’t changing in flight logs
      • Flightlogs where atmospheric pressure changing rapidly
      • Doesn’t seem to be working…
      • Parameter to turn it on and off
      • Off in Copter?
      • Quadplanes and multiple non-drifting things?

UTC2342 -

  • Merged

UTC2345 -

UTC2355 -

  • Is loitering a good thing if you’re breaching the fence?
  • At least one ArduPilot user wanted this enough to get someone to write it
  • Not the default
  • Won’t hurt non-users
  • Has the GCS failsafe part been used/tested?
    • GCS failsafe + RC overrides have to be used
    • Loiter mode and no RC failsafe

UTC2358 -

  • Make hardware define list more readable
  • Only whitespace changes says github delta tool
  • Merged

UTC2359 -

  • Merged!
  • Backing away from operator

UTC0006 - should we use a tool to do auto-whitespace-correction

  • Maybe there’s a button for it in the github repo?
  • Splitting into subdirectories
  • You can write plugins for github to rewrite arbitrarily
    • Maybe Tom can have a look for us?

UTC0006 -

  • FF will review soon
  • Needs to be rebased
  • Important feature
  • Randy would love some review on it
  • Thoughts on:
    • Dijkstra’s only handles 20 fence points
    • Safe points
    • Visibility graphs
    • Arrays have to get bigger and smaller based on number of fence points
      • 20-point limit at the moment
      • memory usage?
  • Tom can review it
  • You can switch between different algorithms
    • OA_TYPE
      • Maybe allow it to switch magically/dynamically?
  • Not merged into AC_Avoidance yet
  • Separate library
  • AC_OA_PathPlanner
  • Not using parameters on dynamically allocated backends
    • Both need a margin?
    • Reuse the margin_max parameter
  • Not a proper frontend/backend split
  • Bendy-ruler gives you a direction
  • Dijkstras gives you a path
  • [10:19 AM] (Channel) davidbuzz: randy , does this bendy and dijkstra have any autotests ?
  • Array of arrays?
  • Fully dynamic?
  • Try to use minimum memory
  • Malloc

UTC0013 -

  • Merged!
  • Whoever plays in here next time should probably remove the # stuff entirely

UTC0025 -

  • This is a rather odd patch
  • Copter-specific
  • Turning on avoidance for everyone!?
  • Chunk more memory used
  • May cause boards to not boot
  • We don’t log the third lane
    • But this enables it
  • We will fall back to the third IMU if the first two fail already
  • Tridge will comment

UTC0030 -

  • User-induced fault for people loading parameter
  • No place to store stuff apart from in parameters
  • Overwriting calibrations is also an issue
  • Setting all parameters and running outside space
  • Volatile / vehicle-specific parameters
  • [10:37 AM] (Channel) davidbuzz: I always thought “number of times a board is booted” is a weird parameter because uploading a parameter file resets it but shouldn’t.
  • Peter will put a PR together to stop MISSION_TOTAL being set

UTC0039 -

  • Deprecate land collective min and replace with collective mid
  • When collective curve feature was added land coll min was left in
  • Users fail to set collective min and it causes issues with ground resonance and dynamic rollovers
  • 3.7 we need spool logic to work properly and this code is involved
    • Chris removed this code in the stuff he flies
  • Risk is that coll-mid is too high then vehicle may not descend
    • E.g. RC failsafe
    • Would vehicle keep rising?
    • Coll-min was supposed to be you-always-land
    • Instructions specifically say to set it up properly
    • Only so much hand-holding you can do
  • Upgrade path?
    • People set mid to hover
      • That goes back to 3.2
    • Is vehicle guaranteed to descend in RC failsafe
  • This will fix ground resonance issue and is worth the required parameter changes
    • Must make sure users are aware of the parameter changes required!

UTC0046 -

  • Add octo-I
  • Unusual config but minimal cost

UTC0049 -

  • Merging methods again
  • Randy wants to talk with MdB about this
    • Touches GPS library
    • Set simulated GPS to NMEA (SIM_GPS_TYPE)
  • Commits need to be split
  • Possibly better protection against bad NMEA streams
  • The == false thing needs to be fixed

UTC0052 -

  • make default for heli for acro to act like a flybar
  • Moved to heli-acro from acro
  • Several users flying this and they like it

UTC0057 -

  • Fix pinout
    • KakuteF7 PWM pinout
  • Default frame type should be same as betaflight
  • Ap_motors_betaflight
    • Type 12
  • At the moment if someone upgrades their vehicle will crash
    • With a betaflight layout they won’t
    • Add line to set motor layout
  • Frame type vs frame class
    • Hal frame type default
    • Kakute f4 went through this transition already
  • New board type with correct motor ordering?
    • Carrying the old one forever seems bad
  • Didn’t match silkscreen because we didn’t have the betaflight motor ordering
  • Force people to redo frame type
  • Tridge had a plan to make this upgrade successfully but can’t think of it right now
  • Put it in release notes and move on?

UTC0108 - Copter and Rover and Plane

  • No real updates on any vehicles

UTC0110 - out of time for this week

  • Did not get to several devcall topic-marked pull requests