Servers by jDrones

Dev Call July 12, 2021

** Issues & Pull Requests**





Attendee count (max): 22

UTC2300 -

  • Compassmot and Plane
  • Useful to be able to do corrections for motor currents in planes
    • Many people are
  • How do we structure people getting those corrections
    • Using post-flight logs and to get corrections for current
      • Works really, really well
    • Do we also want the procedural one as well? didn’t exist when we added the built-in-to-firmware one
  • Scheduler patch needs to be removed
    • Probably before we added scheduler to AP::vehicle()
  • Make this common code rather than duplicated code?
  • Stopping the main loop to run this isn’t great
  • Tridge would be a lot happier if it was shared with Copter
  • Would also be good for Rover
  • This won’t cope with quadplanes
  • How about quadplanes?
    • This only does forward
    • You only get one vector
      • So if you’re only running forward or only running vertical you have an issue
    • Differential thrust also an issue
  • We’d need to support quadplanes with this
    • This is the one that gets affected the most
    • First flight log for quadplanes is a little harder to come by as you need to fly+yaw in qhover
  • Tridge is open to supporting this
    • Preferably getting this into common code
    • Async library rather than synchronous
  • Randy would be happy to do things asynchronous
  • Per-motor compass calibration?
    • Only on a quadcopter
    • And only on quadcopters
      • Really just for skyviper
  • Make the parameters per-battery / per current source
  • In-flight compass fallback to GSF is new-for-4.1 and will stop in-flight bad things happening
  • Must be current-based
    • Throttle-based will be very bad
  • Mask to indicate whether you want to do forward or lifting motors
  • Maintenance burden here is great
  • Compassmot-per-battery for current is most realistic / desirable
  • Is compassmot per compass?
    • Cross-product would suck
  • [9:29 AM] rmackay9: i’m not sure it’s worth the investment in compassmot…
  • [9:30 AM] Peter Hall: So we could do upto four batterys for a single compass with no extra params.

UTC2324 -

  • Allows specification of which frame various things are
  • Fence height limits inc. above terrain makes sense
  • Above-origin isn’t useful?
  • Above-terrain might be a problem?
    • What do we do if we don’t have terrain?
      • Extrapolation?
  • Alt-min is only on some frames?
  • Avoidance will also stop at altitude fence
    • Probably won’t back away from fence
    • Could end up triggering failsafe if you fly forward as it won’t lose height
  • Needs a test suite
  • Reduce scope to just be relative-to-home and MSL
  • General concept of having the parameter is a good one

UTC2340 -

  • Very basic support for the Torqeedo
  • Low-duty-cycle links make good targets for timing-based parsers
  • ] Peter Hall: Can you do two at once?
  • [9:43 AM] Peter Hall: Thinking skid steer
    • No, one serial port, one motor
    • Randy doesn’t quite understand the protocol at the moment
    • Is this good enough to get in?
    • Manufacturer wants an NDA for documentation
      • Wouldn’t be able to open-source the solution?
  • Crc-8 table-based is expensive flash-size wise
    • 240 bytes
    • Use x-modem-style instead
    • Most crcs can be done with table or without
  • Maybe we should remove the larger crc32 table-based implementation?
  • Rts line for gpio?
    • By default, also allow use of gpio for cts

UTC2349 -

  • Pr heli yaw check fix
  • Can be merged after approval from Bill
  • Reset_yaw_target_and_rate and falses?
    • Maybe quadplane would need some?
    • Not a blocker here but does probably need investigating
    • Comment on function for when you want it true or false?
  • Merged
  • Marked for 4.1

UTC2350 -

  • Stop spamming crsf mode/rate messages
  • Merged

UTC2352 -

  • RTL autoland for Copter
    • Petrol Heli really needed it
  • Same as Plane
  • DO_LAND_START support
  • Can only accept it via command long
  • Go to shortest landing sequence OR the DO_LAND_START is in
  • Do users ask for this?
    • They ask for lots of other things
  • SmartRTL wasn’t doing the job
  • Can kind of have rally points with approach sequences
  • Several “safe” paths which vehicles can use at need
  • LUA could be used……
  • This does give predictable behaviour
  • This was essential for the OBC
  • Implementation issues?
    • RTL_AUTOLAND is a bad name?
      • Copied from Plane
    • RTL_AUTOLAND=2 for going to nearest
    • This would switch to auto mode
      • Which Randy doesn’t like
  • Use do-change-mode for coming straight home?
  • New MRTL mode for mission-rtl where we use the mission
  • Multiple missions would be very cool
  • Buzz: A hypothetical new mode called AUTORTL or ARTL or MRTL or something, in plane that gets displayed in the GCS when a plane does a RTL with mission DO_LAND_START , instead of being called “AUTO” does give the plane user more info to understand whats going on.
  • [10:14 AM] Buzz: AUTOLAND?

UTC0012 -

  • Probably want this for 4.1 beta sequence
  • Flight log from Giacomo
  • Diverting in height
  • Double-precision didn’t diverge
    • Single did
  • Stability improvements to EKF
    • Also stopped divergence
  • Good candidate for going back to stables
  • Not sure what sensor data was causing issues
    • Compass-less
      • Was using GSF yaw
    • EKF already starved of information
    • Sensor data not kinematically consistent
    • Only one IMU
  • Mergeable in next fiew days hopefully
  • Henry’s got some new regressions he’ll get replay logs for
    • Particular H743 board with degrading IMUs may be the cause
      • Simply crazy data comes in from the IMU
      • Unflyable on primary IMU, masked off
      • Might be mechanical stress in manufacturing

UTC0017 -

  • Not ready to merge yet
  • Moving origin
  • ArduPilot Extreme project
    • Very high
    • Very fast
    • Very far
  • Tridge can do round-the-world without EKF errors
    • Test suite needs some work
      • Precision issues
    • Vision-based tests are also failing
  • Allows for possibility of user resetting the origin
  • Internal origin shifts once/second
  • Public origin doesn’t change
  • Relative-origin would allow you to change the origin
  • Maybe call he public origin the “ahrs origin”

UTC0027 -

  • Plane update
  • Good reports on autotune
    • New variant coming
      • Probably for next beta
    • Good feedback and flightlogs from people
  • Will almost certainly have the ekf-stability patch in

UTC0028 -

  • Copter update
  • Positive feedback on terrain following
    • Was partially broken
    • Vehicle will not slow down in order to climb fast enough to get over terrain
    • Guided mode will now slow down
      • Auto still needs fixing
  • Use position controller whenever you can rather than using the waypoint controller
    • Don’t use wpnav
    • Use the position controller
    • Precision landing uses wpnav when it shouldn’t for example
      • Rules have changed
  • Heli tail rotor fixed
  • A Matek board fails to boot if ESC telem is in use
  • Gps glitch and emergency yaw reset needs investigating
  • EKF-related items which Paul might get to…
  • Want beta6 out
    • What other important fixes should go in?
    • Auto-terrain-following fixed?
  • Bendy-ruler fixes?
    • Scurves and wpnav at same time so doesn’t work properly
  • Some important fixes in Plane
    • Logging-with-forced-arming
    • Motor-buzzer stuff
  • Waiting for feedback on the fast-uart fixes for gps yaw

UTC0032 -

  • Rover update
  • GPS for yaw is the one users really care about
    • Autonomous mowers

UTC0036 - GSoC update

  • Precland is going well
    • Merged the restructuring of ekf precland
    • Retry is moving along nicely
    • Functionality is fine
    • Code structure needs a bit of work
    • PR is open
    • 4.2 thing
  • Custom firmware server is moving along
    • Covid sucks
    • Good meeting the other day and another this week
    • Basic framework is working

UTC0038 - close

Servers by jDrones