Dev Call Nov 30, 2020

Issues & Pull Requests




1 Like

Attendee count (max): 19

UTC1100 -

  • Points addressed
  • Attach functions to buttons
  • PWM input rather than just logic level
  • Merged!

UTC11106 -

  • RC out over uavcan
  • Still work in progress

UTC1113 -

  • One MAVProxy for multiple vehicles
  • Should change autotest to test with pexpect
  • Needs to work…
  • Nice cleanup

UTC1123 -

  • Needs previous PR to go in - and testing…

UTC1123 -

  • Smaller variables going onto the stack
  • New script to help ensure inadvertent changes to EKF don’t happen
  • Pierre suggested that we could try to have an output version file or similar so we don’t accidentally change the output of the EKF.

UTC1133 -

  • Stops resetting gps type if no vertical velocity
  • Also removes chunks of inhibit-gps stuff from EKF
  • Merged

UTC1134 -

  • Fixes for precision landing because lightware configuration bad
  • Logs show a big improvement in reliability
  • Merged!

UTC1135 -

  • Oblique surveys
  • Recap
    • Hooked into camera ATM to move the mount
    • Randy suggested on AP_Vehicle
      • Doesn’t like the coupling in this PR
  • Tom has a thing where the gimbal tells the vehicle where to go
  • Randy will help Igor structure this
  • Should be minimize-features

UTC1148 -

  • Will be merged after a test flight
  • EKF cleanups

UTC1151 -

  • Save a bunch of memory
  • Follow-up PR to initialise in-flight?
    • What happens if you’re in flight and an allocation fails?
    • Base it on source-selection
  • Merged!
  • Saves 9kB of RAM

UTC1153 -

  • CAN instance assignment
  • Two different GPS with different positions
  • Random in terms of which gets which ID
  • Like compass priority IDs?
  • Or like baro with node parameters / node numbers
  • Paul likes the node ID
    • Configured before putting in airframe
  • Priority allows for sorting between serial / CAN
  • Compass stuff is very, very flexible
    • And very, very confusing
  • Duplicate node numbers across CAN buses?
  • Need comments on the issue!
  • MissionPlanner interface good enough to handle this?
    • Paul thinks if his client had stuck with MissionPlanner things probably would have been OK
  • Magfit results can confuse things?
    • Could output the devids
  • So perhaps we shouldn’t discount the priority-based stuff
  • Priority is a bad name?
    • Is that the order they’re used in or are you specifying which compass1 is, compass2 is etc etc
  • Unique ID for sensor which then gets a priority
  • Rename priority to something else?
  • We don’t currently support twin-gps gps yaw, which is just as well if your GPSs can reverse location

UTC0005 -

  • Geofence not disabled in auto mission during vtol land
  • Tridge will look at this
  • Where’s the Plane / fence thing?
    • Paul will go and dig that up

UTC0006 -

  • Sufficiently deep interrupts could cause stack overflow in AP_Periph
  • Interrupt stack usually at start of RAM
  • Overflows into areas which are not memory
    • Recursive hard fault
  • We have added a low-water mark which we should hit first
    • We throw an internal error
  • Fixes stack sizes
  • Improves stack report
    • Adds ISR to report
  • Nice interface in MissionPlanner to see files
  • Stack pointer too
  • AP_Periph passes a CAN message saying it has hit an internal error
  • Not sent as statustext by default
  • Should we raise an internal error ourself if one of our nodes has an internal error
  • Pass information using can node status rather than statustext?
  • Can look at stack stuff via can node statustext messages now
    • I2c thread was down to 30 bytes of stack left!
    • 20kB of unused memory, so it has been reallocated
  • Stack expanded on h7 and f7
  • Mavlink message for triggering watchdog
    • Can now do it on peripheral nodes

UTC0016 -

  • Needs to pass CI
  • GCS_SEND_TEXT must be used
  • Formatting of message
  • Subsystem should go at start
  • Should remove mission from the message
  • “Only %u items may be uploaded”

UTC0019 -

  • Fixes speed display in OSD
  • Merged!

UTC0020 -

  • Circular limits on pilot desired lean angle
  • Copy of the Copter code
    • Difficult to factor out
  • Qloiter will behave like Copter’s loiter in terms of angle limits
  • Pulling in features from Kris’ branch
  • Merged!

UTC0022 -

  • Another LUA script
  • Copy of the ahrs-source lua script
  • This one changes between different sensors
  • E.g. GPS and wheel encoders
  • Merged!
  • Transition time?
    • Vote counter max constant
    • 2 seconds
    • Randy’s seen 4 or 5
  • More examples coming for source switching!

UTC0027 -

  • Adds a prearm check the the wheel encoders
  • Need to test it on top of tridge’s changes
  • Needs to be rebased

UTC0032 -

  • DJI goggles support for almost all boards
  • Fixes a bunch of typos and makes it all CI’d

UTC0033 -

UTC0036 -

  • Plane update
  • Working with Kris to get features in from tilt-vectored quadplanes
    • New quadplane landing approach and ship-landing stuff is a priority for tridge
    • Nice simulation model for tilt-vectored quadplanes
    • Brandon is available to make the model
    • With a nice model we can do much nicer development and testing on these vehicles
    • Good models means good development
  • S-curves on quadplane to be looked at !

UTC0038 -

  • Copter update
  • No serious 4.0.5 issues
  • If we were to do another release we might bring back the ISR stack changes
  • Bunch of PRs from Amilcar were merged into the 4.0 branch
  • We might have a release manager for a 4.0.6
    • Randy doesn’t want to do the release himself and there’s nothing major pushing it
  • S-curve stuff is actively being worked on
    • 4.1 beta release will go out soon after the s-curve stuff goes in
    • S-curve stuff was a big piece of work
      • Rewrite of the way we do navigation
      • Test plan?
        • SITL
        • Real testing on a quadplane
        • Henry’s TVBS?
      • Landing PR can go in after this
    • Two different s-curves
      • Input shaping
        • Dynamically creating s-curves based on demands
          • takeoffs/landing on quadplanes stuff can come back into Copter
            • But not as part of this PR
      • Waypoint plans
        • This PR
    • Still need to address the units mess
    • Output acceleration translations for different vehicles will come in after this
    • Axis switch?
    • Some new PID types are using ID #0
    • Should pull out more separate PRs
      • Submarine tests
      • Payload place stuff
    • Helicopter tests
    • Huge piece of work
      • Three years of work
    • Need to watch out for division-by-zero
    • Helper maths functions need to cope with zero distances etc
  • How about other stuff that’s based on s-curve?
  • Splines?
    • Current spline implementation with s-curves may not be acceptable for 4.1?
    • We follow the path correctly, but the s-curve stuff is much better
  • Axis switch?
    • We’re using NEU
    • We’ll change to NED
  • Leonard has a plan for degrees vs radians too

UTC0057 -

  • No Rover updates
  • Pivot turns are broken in master….

UTC0100 - off the floor

  • Partner’s call tomorrow
  • AP_Periph stable release in next few days

UTC0102 - close

1 Like