Dev Call June 24 2019 2300 UTC

ArduPilot Conference

Updates & Student reports

Issues & Pull Requests




1 Like

Attendee count (max): 22
Trad-Heli Discussion

  • discussion on tuning Chinook-style tandem helis
  • Discussion of heli in-flight reset issues
    • Autorotation
    • Fly to ground after reset?!
      • Shouldn’t be able to do that
        • But RC control wasn’t lost?
        • No head speed loss
      • No support for pass-through on reset on heli

UTC2301 - GSoC updates

  • Iampete tweaking EKFs
  • Rajat’s modification of the existing ArduCopter interface to do lock-step has proved problematic
    • Bit of a rabbit-hole which has been backed out of
      • The mavlink support was tangling things up
    • Moved to working with simpleflight instead
      • Should be able to use lockstep scheduling a lot easier this way
      • Probably a couple of days work
      • Follow along on the gitter GSoC channel
  • Akshai has been working on the parameter editor interface for MAVProxy
  • [9:13 AM] (Channel) patrickpoirier: well too bad, Just invite yoy guys to take a look at latest Lab Report = We are starting Beta Test

UTC2315 - Copter locking up in flight

  • More Solo users using master
  • Both green cubes and the stock Pixhawk2s
  • Multiple people with multiple vehicles are getting vehicles stopping in flight
    • Motors, logging
    • Rest of vehicle has power
      • Video and CC are still working
    • Watchdog is resetting the vehicle
  • Could be multiple issues
    • Only happening on stock Pixhawk2
    • Not on green cube
    • Maybe the slew limiting is not working as it should
  • Logs don’t show much
  • Lots of logging was added to the watchdog code both before and after
    • Haven’t seen any logs with this in it
  • Unsure if it happens on vehicles without the Solo gimbal
    • Chris’s tlog seems to show exactly the same thing
    • [9:23 AM] To Weekly devcall: We’ve discussed having an IO firmware to allow control in tradhelis.
    • Helis do have cyclic control disarmed
    • FMU must be running
    • Ins gyro cal was set
      • Telemetry was not regained
      • We skip gyro cal on watchdog reset
      • So was it a reset?
  • Really want manual passthrough for heli
    • Probably not relevant in this specific case
  • Logging over mavlink is also Solo-ish code
  • This could have been happening a lot and just put down as a power thing
  • Somebody is reporting they have 4 non-gimballed Solos which are working fine
    • Gimbal is a prime candidate
  • External observations from pilots are useful
    • Video stream information is useful
  • Long-term flying would be nice
  • Brown-out event?
    • We don’t set the brownout voltage
    • We wanted best-effort
    • At very low voltages you can get memory corruption and other badness
    • Now we can handle things a bit better perhaps we could consider setting a brownout voltage

UTC2345 -

  • Merged

UTC2346 -

  • Merged

UTC2347 -

  • Can’t be merged while we’re having stability issues on Solos

UTC2352 -

  • Amilcar - product being released and due-dilligence is being done
  • This is nice work
  • Adding the copyright year bounds apparently reduces your rights rather than increasing them under several modern copyright regimes
  • Adding in a list of contributors is misleading and hard to maintain

UTC2355 -

  • Hoping MdB is happy after failsafe reset
  • Yes!
  • Merged!

UTC2356 -

  • If we’re not using frsky telemetry then don’t use it
  • Race condition on statustexts goes away for those not using frsky
  • Needs testing on hardware
    • Tridge will try it on his small quad

UTC2359 -

  • Extra CPU + bandwidth
    • Can we justify it?
  • Pos was added so we could plot canonical position on map
    • EKF doesn’t log lat/lon
  • Adding a field to add which estimator we are using
  • Need to know how much bandwidth this uses
  • Need to know how much CPU this ues
  • Should we combine POS and VEL?

UTC0007 -

  • Floating point exceptions when changing parameters to obscure values
  • “Well don’t do that then”?
  • Suggested ranges in docs
  • Range-checking for things representable in a float?
  • Runtime checks are going to be painful
  • There will always be parameters you can modify in ways to crash your vehicle
  • Checks on the GCS for these insane parameters?
  • Fuzz-checking?
  • Locking down parameters while armed?
  • This is setting a floating point value to a valid value
    • A derived value is causing the vehicle to crash as it is out-of-range
  • This is actually a false-positive of our FPE code
    • If you turn the enable-FPE in SITL off you’d wobble into the ground
  • Really should be checking for lockups not wobbling into the ground
  • Need a page in our dev wiki explaining how FPEs work in SITL vs real vehicles
  • This value is constrained later
  • Change which faults we raise an exception on?
    • Check outputs not inputs
    • No clear path to get to there from here

UTC0029 -

  • Merged!
  • Need a test to see what happens if the min-loop-time thing is 1ms

UTC0030 -

  • do-land-start and quadplanes

  • Tridge has read but not commented

  • Scripting will be a wonderful answer to this eventually

  • If in qloiter but 2km away from home and you get a failsafe

  • ATM it will come back and land

    • Terrible for this particular user
  • Obvious if you’re within 50m of home

  • Failsafe script would be ideal

    • Complex decision making based on location
  • RTL_AUTOLAND=3 which would not trigger DO_LAND_START in VTOL modes

  • This isn’t a bug - doing exactly what it’s been told to do

    • But not what user wants
  • Lack of situational awareness is the issue here - e.g. qloiter near home should qland but may do-land-start as fixed-wing

  • Throttle issues during RC failsafe is odd

  • DO_LAND_STOP into missions?

UTC0032 -

  • Tridge will test and then merge

UTC0042 -

  • Adding support for GPIO switches
  • Works and tridge flies regularly
  • Lucas wants this moved up to a more common place
  • Tridge doesn’t have time to make this more common
  • Community wants this
  • Allows camera to be used on disco/bebop
  • Tridge will ask Lucas if the generalisation can happen later

UTC0045 - Plane update

  • Tridge would like to do a new beta / release
  • Takeoff PR would be nice

UTC0046 - Copter/Rover

  • Couple of weeks for a new beta
  • Randy is working on object avoidance stuff
    • Before Copter3.7 and Rover3.6 releases
  • Want bendy ruler and dijkstras work with one-dimensional-lidars
  • Memory for proximity library needed
  • Want to eliminate dodge
  • Accelerating very slowly in Dijkstra’s

UTC0051 - Shenzhen conference

  • 3 day conference
  • Similar to last year
  • Big vehicles
  • Quadplanes
  • Ehang was there
  • DJI had big a booth
  • Security is a big thing
    • Police and the like
  • Fewer booths overall, but larger?
    • Industry consolidation?
  • Good to see ArduPilot partners in there
  • University students were great to talk to
  • Lots of tuning questions handled by Leonard
    • Especially on the big ones
  • New boards being produced by partners
    • Tridge will get a ping
  • We should consider next year having a little discussion before the conference about presentations and how they might fit together
    • Unconference presentations may not be appropriate
  • Should we backport nano to stable releases?
  • People are shipping with dev versions now-adays
  • Gimbal manufacturers
    • No direct relationship with ArduPilot
      • Perhaps we should explore this into the future
  • Lots of people wanted direct support from devs
    • Maybe have a “support” or “Q&A” session
    • But this is really what our monthly partner calls are about!
    • Conference is kind of the wrong place to by analysing logs etc etc
    • Face-to-face makes information flow quite fast
  • Hex had a huge, beautiful booth
    • Cube Orange
    • RTF Copter
      • Holding position very well
      • Hexsoon
      • Iris-sized
      • Looked open/hackable

UTC0113 - Close

1 Like