Dev Call Nov 20, 2017 2300 UTC

Contribution of the month



  • 3.5.4 release progress
  • new loiter from Leonard


  • CAN raw
  • Mission upload incomplete - what to do?"


Intent Document Discussion

Attendee count (max): 20

UTC23:02 - Contribution of the month

  • Vote now, result at end of meeting

UTC23:04am - tridge and Plane

  • RealFlight and models

  • Lorbass has been doing some great work on creating models

  • Starting working on quadplane and other models

  • Makes a huge difference having a model which is realistic

  • Modelled off a real aircraft

  • Tridge loaded parameter from real aircraft and it flies using those!


  • Online form to enable to capability to talk to ArduPilot


  • Tridge would like feedback on whether this works

  • Should be free

  • SITL models repo


  • Room for XPlane, JSBSim etc etc models in here

  • Would like deep-stall landing models for MdB’s stuff

  • Better than JSBSim?

  • JSBSim has a lot of potential

  • Harder to create a model in JSBSim vs RealFlight

  • Easy to play in RealFlight

  • E.g. trying different airfoils, props and geometry for TVBS

  • Best model, probably

  • XPlane uses blade element theory, which should be pretty good from the physics perspective. It has a maximum number of elements though, so for complex shapes some linearization errors will be present.

  • Tridge: main issue is frame rate is quite slow

  • Working on a plugin to make it faster

  • But still order of magnitude slower than realflight

  • Works OK on big vehicles like Cessnas

  • No good on multicopters

  • Limitation on motors, too!

  • Can’t do an octoquad plane

  • Should be able to get lock-step scheduling in XPlane

  • AirSim?

  • Graphics are spectacular

  • very , very slow

  • Framerate was a few Hz

  • Physics engines are very simple

  • Like the built-in-SITL one, really

  • Does vision stuff which is great

  • Apparently getting faster

  • We don’t support it because tridge doesn’t have a fast-enough machine!

  • Blender can export vehicles into RealFlight

  • Elevons rotating around right axis is current problem

  • Losing some axis orientation stuff in pivots?

  • Tride is using Blender to attempt to model using free tools

  • Turbulence model vs wind model

  • Seems to over-do the turbulence when raising wind

  • But it is settable…

  • Much more realistic winds

  • Small bugfixes in Plane over the last week

  • Tailsitter transitions in auto have been fixed

  • New control algorithms coming from Paul for tailsitters

  • Plane of rotors rather than plane of autopilot

  • Doesn’t control pitch attitude of wing

  • Controls pitch attitude of rotors

  • Lets plane swing in the wind

UTC23:19 - Randy and Copter

  • Compass fixes in latest -rc
  • Will make official version in next few days if nothing goes wrong
  • Also the serial 5 patches went in
  • Working with Leonard on new loiter algorithm
  • Pretty good
  • Responsive in roll/pitch
  • Response when you let go of the sticks is still problematic
  • Weaker than it used to be
  • May do a correction / move around which is no good
  • Will be PR’d soon
  • Not enough feedback from people so far!
  • Loiter is very commonly used, we need to be careful with it
  • Has to work well
  • Copter flight-mode objects
  • Just fiddling about moving functions into the classes

UTC23:24 - Randy and Rover

  • Rover release coming
  • Bunch of PRs from Pierre
  • Much better than what’s out there
  • Instant reversing issue when pivot-turning is a problem
  • Users don’t like it….
  • Before release or after is not clear yet

UTC23:25 - request to get MAVProxy and CUAV changes in

  • Don’t want to lose the changes that were so helpful during the challenge
  • Fixes it for newer versions of wx3
  • Breaks it for older wx implementations
  • Tridge merged it along with some other bits and pieces

UTC23:33 - CAN Raw with Tom

  • Random other proprietary can products in-house
  • Want to convert to uavcan
  • Eugene’s infrastructure should allow lots of CANs quite well
  • One enumeration is a problem
  • Need to add an option for different enums
  • No UAVcan HW at the moment
  • Similar problem to other people
  • Allowing arbitrary CANs would allow for easier integration
  • Would be nice to have J1939 CAN based servos
  • RAW Can allows you to switch from UAVCan to raw can which allows you to write directly to the port
  • Stick in UAVcan or bust scenario ATM
  • if you look in the ardupilot/canbus gitter channel, someone has tested CANOpen esc’s successfully.

UTC23:41 - incomplete mission upload with Francisco

  • What should happen if a mission is not fully uploaded?
  • Constantly updating mission
  • Need ArduPilot to know if the mission was completely uploaded or not
  • E.g. you’re in guided mode and something happens
  • Radio failsafe
  • New code in AP which changes mode to auto instead of RTL
  • But only want to change to auto if mission upload complete
  • RC failsafe and do-land start could start incomplete landing sequence at the moment
  • Could take some other action instead
  • Perhaps if next waypoint isn’t available we do something instead?
  • Could currently end up in a loop in Plane with the do-land-start stuff!
  • Copter could reject the auto mode change based on whether you are uploading a mission at the time
  • Mission extension in flight is something people do
  • Add flag to mission upload
  • Atomic mission uploads
  • Consume RAM
  • MdB: 512 item missions
  • Still only 8k of memory!
  • Non-atomic upload if you don’t have enough memory?
  • Would need extra mavlink messages
  • Being able to tell what mission is on the autopilot would be good, too!
  • Need to be able to fetch from the vehicle what it is actually flying
  • Current mission protocol wouldn’t allow you to atomically upload a mission while retrieving the old one
  • Transaction ids?
  • fighting over setting missions
  • E.g. CC and GCS
  • Should we start opening up our options based on running Linux?

10:55 AM] (Channel) So, Peter now we can have a look at the PR that extends the max wp per mission on linux boards ??
[10:55 AM] @Luis sounds like Tom would love to get that merged :wink:
[10:56 AM] @Peter, everyone that tries to create a circular survey on MP is immediately bitten by the max number of waypoints
[10:57 AM] @Luis: can we improve the circle mission item and get MP to use that?
[10:57 AM] I thought Copters really sucked control-wise with that sort of mission, too?
[10:58 AM] not likely @Peter, Also Don just released a vertical structure scan mission creation that also is unusable on ArduPilot, due to the large number of WP that it generates

  • We should use a flag in the short term and come back to this later
  • MIS_OPTIONS bit which says don’t allow entering of auto mode if we are part-way through an upload
  • Set the mission count?
  • Rally point counts are a nightmare -mdb

[interleaved discussion in mumble text]
TP: I’m just saying we have a nasty habit of working out the nitty gritty details in this call where sometimes it’s best to make overall decisions in this call, then have other more personal “off-line” discussions to figure out the fine details
[10:58 AM] @Tom: I like the details. I learn stuff.
[11:00 AM] TP: well, as long as we have 2 hour long marathon calls working out the low-level details then we won’t get any higher level people in the discussions from other companies. We waste too much time working on the fine details with 20+ people in a room just listening. I think we could get more people here and have more wider-variety of topics if we didn’t spend 30 minutes on given a topic that some people don’t care. They tune-out and leave
[11:01 AM] @Tom: you want something between our dev call and our partner calls?
[11:02 AM] @peter yes. the partner’s call is usually just us reporting a status. There’s a little bit of discussion but… yeah… a weekly thing with a mixed audience would be nice

UTC00:09 -

  • Need to preserve stream rate adjustments based on radio status
  • Should be able to do this with very little memory consumed

UTC00:33 - docker support

UTC00:42 - update home on arm

  • Most people don’t arm the vehicle with the safety switch not pressed
  • MdB uses the safety switch as a “go” button
  • We update home continuously when armed but switch not pressed
  • Soft-armed true requires safety switch to be pressed
  • Tridge will merge
  • To get into this scenario you have to disable to safety-button check
  • Integrator windup is prevented with current soft-armed logic
  • Takeoff windup prevention is elsewhere though
  • disabled/enabled/doesn’t exist
  • @MdB have you considered having a “go” AP_Button rather than overloading the safety switch functionality? :slight_smile:
  • [11:44 AM] Peter: sorta but the motor inhibit is a bonus

[b]UTC00:46 -

  • Would be nice to get this one merged
  • Sbus out on PixRacer
  • Make serial manager a singleton?
  • Can do it from board code then
  • Tridge has a PR which adds that (the Vz-servo patch)

UTC01:05 - Contribution of the Month

  • Bill Guyer and Chris Olson are this (two) month’s winners - Thank you and congratulations!

UTC01:05 - Landing Issues (Plane touch and goes Gary Mortimer)

  • Would land and it would crab
  • When we touch the ground we should remove the wind compensation
  • Ignore wind and straighten up
  • We already have this
  • Separate servo wheel for wheel and rudder at all altitudes
  • Also using rudder to point down runway like real plane
  • Mk1 eyeballs on real aircraft make this work
  • Lidar
  • Weight-on-wheel sensors would make this easier

Hi Craig, Thanks for the great news!! By the way my last name is spelt Geyer.

Thanks again!