Dev Call Sept 28 2020

Issues & Pull Requests




1 Like

Attendee count (max): 22

UTC1100 -

  • frsky rescaling
  • Merged

UTC1105 -

  • Remove two-stage requirement for wind failsafe
  • Users are missing they need to set two things
  • Wind_max only used to affect stuff in ahrs
    • Now it will affect airspeed, setting sensor to unhealthy
  • User log shows airspeed as running very high
    • Sensor was actually really unhealthy
    • Never switched sensor off
  • WIND_MAX was only supposed to be used in AHRS
    • Never for the sensor health
  • Why didn’t it work already?
    • Why didn’t the wind_max code inside airspeed_estimate catch this and adjust the airspeed estimate based on the wind_max
    • Get_wind_max is only used in dcm and this place in airspeed
  • Quite a departure from current usage of WIND_MAX
  • This wouldn’t fix the real bug
    • The airspeed_estimate function isn’t clipping at WIND_MAX?
  • This is probably custom code
    • Lacking the recent patch by Peter Barker that fixes it
  • Big gust of wind coming over a hill
    • Might turn sensor off
    • Does it turn it back on again?
      • Second bit allows for that…
  • E55057ad5ddb1fc74c54bd13c264fc54e2c47719
  • Need to understand the problem before patching it
  • Users more tolerant of “sensor failed so we crashed” than “we thought it failed so we stopped using it and the vehicle crashed”
  • Might have a codepath which doesn’t call airspeed_estimate?

UTC1121 -

  • Change shown HRS view once transition complete
  • Changes logging and what is sent to GCS
  • ATM As soon as you switch modes we change the view
  • Instead we wait until you finish transition - when you switch controllers
  • ATM Euler angles are nonsense until you finish transition
  • Bad on the ground?
    • Switch modes wouldn’t switch views
    • There’s a timeout which does seem to trigger
  • Actually works on angle_wait, not the transition-complete
  • Will change it to switch immediately when disarmed
  • Nice change, needs some patching

UTC1128 -

  • Motor state being set incorrectly ATM
  • Is_manual_throttle vs is_auto_throttle / copter vs plane
  • Needs some comments…
  • Throttle mix currently incorrect
  • Throttle-mix-man, throttle-mix-auto
    • Big angle error, should we accelerate more?
    • Attitude vs vertical control

UTC1133 -

  • Do lots of niceties before we reboot the
  • We might not need the zero-rc-outputs any more
    • ChibiOS might fix it
  • Do we need to test this on hardware with a logic analyzer?
  • Slight race condition
    • Rc output thread will try to overwrite it again?
  • Do the zero-rc-outputs on all vehicles
    • Flash it onto a Plane and make sure servos behave?
  • Separate PR for changing to zeroing RC outputs on all vehicles
  • Uavcan ESCs are going to be a problem
  • Can be merged after the int vs int8 is fixed
  • Need to raise an issue for constraining uavcan servo outputs
    • Volz?
    • Thwacking to endpoints
      • Should be constrained based on fancy servo internal configuration…
  • This PR can be merged after the int fix and a check of short-pulse on a Saleae

UTC1154 -

  • Stop processing when RC failsafe is on
  • There are race conditions with has_valid_input
  • Change API so that get_pwm tells you whether you got good RC data
  • Rc8.get_pwm
  • Multiple return values in LUA using a tuple
  • A boolean function which fetches an array of RC inputs would be safe
  • Won’t be able to fetch raw pwm values with such an API
  • Just need to add a check in the get_pwm call

UTC0008 -

  • In auto, disarmed in ground
  • Want vehicle to take off and fly with a single mavlink command
  • Bypass throttle at takeoff
  • Taking off without origin was a problem…
    • Don’t run mission commands unless we have a valid origin
  • Spool-up is a problem for big Copters when doing something like a guided-takeoff
  • Waypoint navigation controller
    • If you increase the spool-up time it gets bad…
    • Little leap off ground
    • Any auto takeoff….
  • Mixed-case enumeration values are now acceptable
  • What would Copter do if you armed in auto and the current waypoint is not a takeoff?
    • Failure in init
  • Will need to be on a takeoff waypoint or landed

UTC0022 -

  • Leonard and Bill seem to be happy with this
  • Works pretty well in simulation for Copters and Helis
  • Works well on a Plane in practise
  • Can prevent oscillations on ground for helis
  • If you push D or P too high it can prevent the oscillation
  • Even in tuning you can use this
    • Logging
    • Still oscillates just with lower amplitude
  • _STC rather than _STAU
  • Should we default this on Copters to some value?
    • Might save vehicles
    • 40 or 50 on an Iris is about right
  • Can tell from a log what this value should be
  • The RATE message shows the slew rate we’re actually demanding in a flight
    • PR shows logs from a racing quad which should show extreme
    • Also one from a heavy-lift Copter
  • Two extra parameters per PID
    • Time-constant of 1 is probably fine for everyone
      • So remove the parameter?
      • 1 is 1 second for restoring gains
  • GCS warning when the slew rate activates
    • Takeoffs on flying wings on Planes trigger this
    • Landing on helis and takeoffs on helis
  • Can use this for sharper tunes in normal flight, but can still have controlled takeoffs
    • No wind estimate and speed scalar way off
  • If we drive outputs faster they can physically move you can also get bad effects, we don’t protect against it
  • Two videos showing Copter and Heli benefitting from this
  • To Be Continued….

UTC0038 -

  • Merged!
  • ICE engine / RC bug

UTC0045 -

  • Identify ArduPilot binary without running it
  • Scanning the binary….
  • Compiler might include the symbol multiple times?
  • Need to make sure the symbol only entered once
  • Scanning is fine
  • Need a script which extracts the information in Tools/scripts
    • So we can test it

UTC0055 -

  • Some quadplanes need ground effect compensation….
    • 12m error on baro on landing
      • Bounce!
  • Needs a rebase
  • On by default rather than having a bit?
    • Yep
  • Option to turn it off
  • Different method on different vehicles to do the same thing

UTC0058 -

  • No link to the mavlink PR?
  • Were the messages moved accurately from ardupilotmega.xml to common.xml?
  • Moving them into common.xml mean open-season?
  • EFI message problems….
  • What’s in it for us, moving these messages up?
    • QGC
    • Upstream have already made these changes
  • Sort-and-diff the id/checksum
  • Deprecation of the data-streams messages
    • Deprecated for 5 years
  • [11:08 AM] Peter Hall: Do we have many users on QGC? Just mac and Android people?
  • [11:08 AM] brunoolivieri:
  • Do we have many users on QGC? Just mac and Android people?
  • @Peter Hall Android is quite useful on the field…
  • [11:09 AM] joshanne: Can confirm we’re users of QGC
  • [11:09 AM] Peter Hall: MP is comming to android soon.
  • [11:09 AM] joshanne: many youtube channels I’ve come across are also users of QGC
  • [11:10 AM] rmackay9: There are quite a few professional users of QGC. I think QGC is less functional but prettier than MP…
  • [11:11 AM] Peter Hall: Easier for people to skin too…
  • MAV_SENSOR_ conflict
    • Equivalent rotation?
    • Completeness test
    • Need to remove the duplicate

UTC1107 -

  • Remove unused macro
  • Merged!

UTC1117 -

  • Limit the rate at which we accept interrupts
  • 1MHz interrupts….
  • Pushing Write_Err to the GCS might be quite nice
  • Sensor will stop getting data so can tell from logs what died
  • Might drop a send_text in as well
  • Need to start a new stable release
  • … and test the notification system
    • Don’t want to get people worried?
    • Mitigation?
      • Stop using interrupt-driven input like RPM encoder or whatnot?
  • Small stable release?
  • Will merge the fix and ask user to ensure it is fixed
    • Tridge will work with Stephen

UTC1126 -

  • Randy to look at

UTC1134 -

  • Allow sending positive/negative values to UAVCAN ESCs
  • Will need a parameter to enable this
    • Setting a bit might make your motors go to half-throttle….
    • Make it reboot-required?
  • Blheli has three separate ranges
    • Reversed
    • Forward
    • Don’t turn
  • CAN doesn’t have a “don’t turn” value
    • It does….
    • Use DISARM_MASK….
  • UAVCAN parameters?
  • Do we really want to add more very-similar parameters or get some common ones?

UTC1140 - Plane update

  • Just the presence of the above issues
  • And another stable release coming up

UTC1147 -

  • Rover update
  • Can’t get log list?
    • No steps to reproduce?

UTC1147 - Copter update

  • No issues to report on 4.0.4….
  • Vibration failsafe has been saving vehicles
  • Randy’s been flying into and out of vehicles - see videos
    • GPS speed accuracy to determine switching
    • [11:49 AM] Peter Hall: I often fly into buildings
    • [11:49 AM] Peter Hall: also trees

UTC1150 - close

1 Like