Dev Call Feb 24, 2020

Issues & Pull Requests

Compass Calibration issues




1 Like

Attendee count (max): 21

UTC1102 -

  • Fix Plane RTL behaviour
  • Niche set of circumstances which could create a bad RTL behaviour
  • Rtl-autoland and do-land-start and mission-restart so it resets mission
    • And does not fly home
  • Calling function of start/resume to force resume
  • MIS_RESTART is the key - doesn’t get tested too much
  • Merged and marked for backporting
  • Is this vehicle-specific?
    • [10:07 AM] To Weekly devcall: I must say I’m not a great fan of poking state into the library - but this fixes a bug.
    • [10:07 AM] To Weekly devcall: I’d rather liked to have seen an autotest for this so that if Randy decided to fix it in some other way that we’d have some coverage.
  • Matt will have a think about whether this can be fixed in a different way, too

UTC1108 -

  • Resolves a regression in 4.0
  • Landing gear always retracts
    • This was deliberate
  • Want an options parameter to do this?
    • This is that
  • [10:11 AM] To Weekly devcall: <- those accessors could instead be “option_set(AP_Landing::Option option) const”
  • [10:11 AM] To Weekly devcall: By the time you’ve got 16 of the damned things that would be a lot better :slight_smile:
  • Only done on Copter for the time being
  • Should be done on Plane, at the moment we will just ignore the parameter
  • Tell the library what we are doing rather than telling it what to do?
  • Edge-triggered
    • Landing_gear_should_be_deployed isn’t edge-triggered
    • Should move to being edge-triggered
  • Probably need the same fix in Plane

UTC1120 -

  • Couple of complaints about landing detector not working
  • Uses filtered acceleration where previously it was not
  • Four or five conditions for landing to be considered accomplished
    • Attitude control’s vs throttle
      • High-powered Copters have trouble with coming down / maintaining low throttle
      • Attitude-controller-adjusted-throttle
      • Half of hover throttle required
  • Quadplane equivalent:
  • Vibration bug in SITL in master?
    • SIM_ACCEL parameters incorrect
  • “Blend” is historical?!
    • Just uses a single accelerometer?
    • Maybe non-isolated IMUs stopped us using blended things as actual blended values

UTC1143 -

  • Cleanup and extend ramtron driver
  • Tridge reviewed
  • Probably doesn’t explain lost parameters
  • Fujitsu have a different layout of their registers vs cyprus
  • Tridge has an open PR for not doing fallback
    • Lost parameter problems
    • Mixture of storage-on-SD and storage-on-RAMTRON is bad
    • There may be boards out there with dead ramtrons
      • If we apply the change then we will break these boards
  • This PR added a new define for not falling back to SD card
  • Would need a “no ramtron” build to make the boards work
  • User had a relatively reproducible case
    • Specific powering and peripherals
    • Power vulnerability
  • [10:47 AM] To Weekly devcall: We could potentially add a epoch time into a parameter - that would let us catch a case like this in some circumstances.
  • Retries on reads and a few other things didn’t fix the problem
  • These patches add some good things
  • Tridge will continue to review this with Nick
  • Jaaky has boards where the FRAM doesn’t work at all and he’s using the SD card fallback
  • Tridge made a patch to stop all writes to the RAMTRON
    • Parameters were still reset!
    • Read destructive?
    • Voltage spike means complete reset
      • Not fixable in software, perhaps
  • [10:47 AM] To Weekly devcall: We could potentially add a epoch time into a parameter - that would let us catch a case like this in some circumstances.
  • Probably add several bits in soon (inc crc) but the root cause is still unknown

UTC1156 -

  • Logs basic information about each mavlink link
  • merged

UTC0000 -

  • Fix crash detection
  • Merged!
  • And general agreement that having an action-none would be good

UTC0002 -

  • Smoother descents based on rangefinder
  • Handles glitchy rangefinders better
  • Using inertial?
    • This hasn’t been tested on a real vehicle
    • Same logic has been flown a lot, just not this implementation
  • Merged!
  • Should add for upward-facing too

UTC0009 -

  • Retry neopixel until success

  • Fixed possibly better in another PR

  • Problem where some pixel’s data doesn’t get through at all

    • Serial / DMA issues?
    • Rangefinders shouldn’t matter
  • DMA memory allocation issue?

    • CubeOrange shouldn’t care
  • Would be nice to get to the bottom of the lost update problem rather than papering over it

  • Peter Hall will dig further on the issue

  • But we’ll merge this anyway

UTC0016 -

  • Are we OK with scripts setting arbitrary lat/lon?
  • Fences…
  • Be careful with your scripts
    • Don’t put dodgy scripts on your vehicle….
  • Takeoff and velocity control coming
  • Assert for vehicle type at top of script

UTC0018 -

  • Receivers can take a long time to boot
    • Already have a race condition
  • The txmod stuff on rfd gear
    • Another race condition
  • Adding complexity for no advantage
  • Arming vs disarming
  • Armed-and-flying
    • First RC packet disarm shouldn’t disarm?
  • Disarmed on ground and first packet -> ignore
  • Interlock and parachute release
  • Default setup on most transmitters will complain if switches are not in correct default location
    • No correct position for an arming switch
  • Fix Throttle failsafe instead?
  • Narrow scope to specific switches?
    • Fill in the act-on-first-rc?
  • Only so many use cases we can solve in ArduPilot
    • Can only prevent so many issues
  • User story is getting really really complicated
  • Peter will put in some logic around arm/disarm

UTC0044 -

  • Merged
  • More consistent across vehicles

UTC0049 -

  • Merged!

UTC0053 -

  • Make it a uint8_t in the routing table

UTC0059 -

  • Randy doesn’t mind if we change behaviour
  • We can move to this!
  • Default to zero?
  • Default rate
  • Persistence?
    • We rely on persistence in Plane for post-watchdog reset stuff
    • Eeprom write issues?
    • Writing the USB port in flight while USB connected
      • Paying the cost on Plane already
    • MdB and Randy don’t like GCS
  • Agreement that we can persist parameters on all vehicles and that then GCSs can stop poking the vehicle
  • A 1Hz output on all channels is fine for the time being

UTC0015 -

  • We are not maintaining 3.6
    • Many safety-critical fixes are not in 3.6, users should move forward

UTC0016 -

  • Option to rewind a number of waypoints on resume
  • Approach has changed a lot
    • Rewind a number of metres
      • Mission-specific sort of a number?
    • Steps back through the history until it finds two waypoints where the rewind distance is satisfied
  • [12:17 PM] (Channel) MdB: this honestly is starting to sound much more like a script thing rather then AP_Mission :confused:
  • [12:17 PM] (Channel) MdB: just my opionion :smiley:
  • [12:17 PM] (Channel) MdB: sounds complicated :smiley:
  • Make it a “DO” command
  • Don’t go behind the DO in the mission
  • Command index for n waypoints behind instead of a complete history?
  • If you are flying very long legs then you can resume mid-way
  • If it is negative it is in metres if positive a boolean?!
    • No parameter any more - only becomes active on receiving the do-set
  • [12:28 PM] To Weekly devcall: I’m guessing this doesn’t work with spline waypoints?
    • Anything with a lat/lon
    • Doesn’t do splines
  • Andy will go away and make some changes and bring it back

UTC0035 -

  • Fix printf error
  • Already fixed in maser, closed

UTC0038 -

UTC0038 - compass calibration issues?

UTC0041 - Plane update

  • New beta
  • Beta is similar to the Copter one
  • Keeping the two releases aligned
  • Loiter-turns bug and a TECS reset added
  • Conflicting parameters
    • Guided mode as hover instead of circle
      • We have a parameter to disambiguate
    • LOITER_TURNS while hovering is rather confusing
    • Conflict led to not completing loiter-turns
  • Max climb rate fixed wing very low
    • Exposed issue where it would descend again before climbing to first waypoint

UTC0043 -

  • 4.0.3rc1 out there
  • Most improvements are good
    • Zigzag and do-set-servo changes work
  • Blheli passthrough on CubeOrange is still a problem for some users
    • PH thought the baudrates were correct
    • It isn’t reliable for tridge
      • Worked reliably while logic analyser was attached
      • Some sort of noise floor or pull-up issue
        • Analogue domain bug
      • Maybe something in the filtering in the H7?
  • GPS blending issue?
    • Two of them and GPS blending we get an internal error which is a contrain-nan issue
    • Randy will do this patch
  • Bunch of PRs fixing other issues
  • WPNAV_ALT_MIN next one to look at

UTC0048 -

  • Nothing for Rover

UTC0049 - close

1 Like