Dev Call Nov 4, 2019 2300 UTC

Issues & Pull Requests




Canberra Conference

1 Like

Attendee count (max): 21

UTC1101 - Daylight saving

  • Apparently MdB doesn’t suffer from this

UTC1102 -

  • We’ll be going with internal error loop here

UTC1107 -

  • There are two versions of this
    • One more elaborate than the other
  • Do we want to introduce a base class for the EKFs?
    • Save a bunch of flash space
    • Would Paul be OK with it?
    • We should be trying to replace EK2 with EK3 at some stage
      • Estimating gyro scale factors problem
        • Accel biases in EK3 is more useful
    • We don’t have a plan for getting there
      • Largely testing
      • Tested on vehicles like Solo
        • And a few others
      • Confidence to make the switch
        • And time to dive into failures and fix issues
    • Tridge is using EKF3 a bit
    • Need to discuss with Paul
    • EK3 has GPS yaw
    • EK2 has vision input, EK3 doesn’t
      • This definitely needs fixing
        • Goes nuts over a minute-and-a-half
  • We’ll try the parent class thing
  • Mark can go with the simpler solution for the time being
  • So the other PR?
    • Can be merged

UTC1117 -

  • Peter to test once more in MAVProxy then it can be merged

UTC1131 -

  • Another watchdog reset
  • Older code without the storm mitigation
  • User will be upgrading
  • Randy has tried to advertise to people that they should really, really upgrade
  • We haven’t seen the problem on 3.6.11
  • Tell all your friends to use 3.6.11!

UTC1133 -

  • Motor tilt when disarmed
  • Vectored-yaw vehicles
  • Convergence user didn’t think things were working
  • Rudder-stick movements should move the motors?
  • Tridge isn’t a fan
    • Imagine disarm in flight?
    • Motor will also be off….
  • Shouldn’t they already tilt when disarmed?
    • Maybe only if used for roll/pitch
    • Possibly not for yaw?
    • Would let you test your yaw range / binding on servos
    • HSV_MAN thing on Heli is relevant
  • We have a relatively nasty hook into AP_Motors to fix the assumption that thrust is straight down
  • Mark to test whether yaw works appropriately
  • Maybe make this an option?
    • Wait for someone to complain
  • Max-angle issue?
  • If you’re doing arming with rudder then motors will all be tilted over
    • And mot-spin-armed might mean they start spinning while deflected a lot
    • … and vehicle may actually choose to just go.
  • Maybe add a bitmask?
  • Need to test this
    • Especially with rudder arming
  • On heli we reset everything when arming occurs
  • Maybe add a manual-pass-through mode instead?
    • Like HSV_MAN
      • Which is magically unset
  • Using PREFLIGHT_CALIBRATION is probably the correct thing to do

UTC1149 -

  • Having a last-valid-values method on RC_Channels would be nice
    • Make get_radio_in return the last valid value
    • And have another method to get the current value
  • This is a problem when receivers give bind values rather than (e.g.) S.BUS error bit set
  • Tridge wants the better fix for the next Plane fix
  • Tridge or Peter will create a PR to do this properly before next week

UTC0002 -

  • Running out of servo options
  • Proposal to make them int16
  • Conversion code required
  • Peter Hall has said he’ll try and do the conversion

UTC0010 -

  • Namespace changed, should be NFC
  • Randy will keep going through it and give it a review

UTC0011 -

  • Testing is a problem
  • Peter will go off and write a simulator

UTC0021 -

  • merged

UTC0023 -

  • Merged
  • Sub now advertises it can do compass calibration

UTC0028 -

  • Merged
  • Some dissension on correct way to cast

UTC0030 -

  • Ensure consistent compass ordering
  • Discussion was had on weekend
  • Devices aren’t always enumerated in same order
    • Problem for things like compass offsets, gps position on vehicle, …
  • Idea:
    • Devids
    • Not indexes
  • What happens if it doesn’t exist?
    • Must have been stored in eeprom
    • Calibrated with sensor that no longer exists
    • Prearm failure
    • Force-arm would consider sensor unhealthy
  • Apply semantics to all sensors
  • GCS authors would need to add support for this
  • Compass-cal would set these IDs
  • How about in-flight compass-calibration?
    • How about stick compass-calibration?
  • See also

UTC0042 -

    • RTCM injection
  • Tridge is still working on this
  • Devcall removed
  • More testing required
    • Need to ensure RTCM is actually used rather than being received

UTC0043 -

  • SITL fix ADSB
  • Merged!

UTC0048 -

  • Peter to add length to the API
  • And to add same to the send_message API

UTC0057 -

  • Message extension is in upstream mavlink
  • PR to add it into our mavlink
  • This is the PR which uses that extension
  • The mitigation thing has a problem
    • Entry 0 should be UNKNOWN, not NONE
    • This is a safety issue
  • Peter Hall can push branches into our mavlink repository to get his PRs passing

UTC0103 -

  • Add NMEA input library
  • Not used for GPS
    • But could be
  • Adds more lines of code
  • Want to make a parser rather than a class thing
    • Two decoders in GPS…
      • Detecter and main one

UTC0007 -

  • Manual throttle for various Plane Q-modes
  • Is this the best way we could do forward throttle?
    • Extra slider on transmitter to set amount of forward throttle
    • Whereas stick controls pitch
    • Another degree of control to worry about now…
    • But you can pitch down and still climb

UTC0014 - Plane update

  • Want to do another release soon
  • Matek thing
  • The RC channel pulse problem
  • Takeoff fix is already in 4.0.0
  • People seem to be happy with takeoff
  • One report of a reboot using sdp33 airspeed sensor
    • Watchdog trigger + locking error?
    • Waiting for logs

UTC0015 - Copter update

UTC0032 - Rover updates

UTC0033 - closed

1 Like