Dev Call Oct 5, 2020

**Issues & Pull Requests"




1 Like

Attendee count 20 (max):

UTC1100 -

  • Backporting fixes
  • Rangefinder code should be using a median filter from a library rather than doing its own
  • 4.0.5 hasn’t been put out yet
  • Prearm check bit is nice
  • Tridge could bring these into Plane as well
  • Looks harmless
  • Will leave this one for later as we want to get 4.0.5 out and nothing in here is critical enough to delay 4.0.5
  • 4.0.5?
    • Watchdog reset
      • No resolution as yet
      • Problem is on 4.0.4 so 4.0.5 will be no worse
      • Race condition in DMA on UART driver
      • 4.0.4 is relatively young
        • Perhaps the watchdog came in in 4.0.4?
      • Need to get a reproduction
  • If we do another rc we might pull this one in

UTC1116 -

  • Add precision to vertical airspeed
  • Add extra digits at lower vertical speed

UTC1117 -

  • Fx prearm on M9
  • Arming failure because its failed to set up tmode / sol speed
  • Works on f9 but not f9
  • Need to distinguish on software version string?!
    • Nasty but no other way?
    • MON_SVINFO dropped?!
      • No more capabilities message
      • M8, m6 use space-vehicle-info packet
  • Merged

UTC1121 -

  • Log vibe as instance message
  • 10Hz
  • Fixes to multi-instance graphing available in mavproxy
  • PID logs in Copter
  • Can be merged once CI passes

UTC1127 -

  • Run vibration monitoring on all instances
  • CPU load?
    • E.g. CubeBlack (F4, three IMUs)
  • On F4 run on selected lane on primary ekf instance?
    • Run F7 and H7 run vibe loggers on all cores
  • Tricky ifdef?
  • Still need to log all three so we get clipping
  • How much stuff do we change on memory class?
    • Log buffer size
    • Should use it in SmartRTL
    • Randy thinks we should just use “300 and above” rather than all the other levels
    • The “rtl pause” in smartrtl does need to be measured
    • Orange (H7) has a lot of RAM @1000kB
      • Across 5 regions
        • Weird memory layout
        • 1055kB

UTC1141 -

  • Add an EKF prearm_check call rather than a failure-reason etc
  • Part of wider change for mixed gps/non-gps navigation
  • DCM shouldn’t change in behaviour

UTC1154 -

  • Always emit gps2_raw if it is configured
  • Little change to make then can be merged
  • A few cleanups to do in the GCS

UTC1156 -

  • 3.6 behaviour where GCS failure would not invoke for pilot-controlled modes
  • This maintains 3.6 behaviour moving to 4.0
  • Default changed going from 3.6 to 4.0
  • Change the default behaviour for everyone
    • So by default it doesn’t change mode if you’re in a pilot controlled mode
    • Massive change in behaviour?!
    • Consistency across vehicles would be good!
    • Expected default behaviour for a GCS failsafe should be to not change mode if you are in a pilot-controlled mode
    • People with RC failsafe turned off will get a surprise when their vehicle doesn’t failsafe at all
    • Surprising number of large vehicles flying out there without RC failsafe set
      • 3.6 to 4.0
  • Add complex behaviour when user has neither check enabled
    • Add a prearm check instead?
      • Easier to understand
    • Arming check will ensure you have one or the other
    • Based on the “parameters” check bit

UTC0010 -

  • T265 to ignore position and speed after reset
  • Camera can “run away” if it gets confused
  • Happens quite often
  • Glitch protection in scripts
  • Speeds >20m/s
    • Position reset bit gets set
  • Why isn’t this being caught by the EKF innovation gates?!
  • Maybe it’s rejecting velocity but is fusing position?
  • Reasonable bandaid?
  • We do have glitches in simulated vicon
  • merged!

UTC0017 -

  • Correct compilation with various features disabled
  • Merged!

UTC0024 -

  • Add a type
  • Separate things e.g. BTN1_PWM_MIN
  • Tridge likes the addition of the function
  • Works fine with the switch
  • Concerns with interrupt positioning
  • Tridge would prefer to retain the existing timer-based boolean-switch stuff
    • Aux2 and aux5 can’t have interrupts attached
  • Might hit the too-many

UTC0044 -

  • Remove vi comment then can be merged

UTC0046 -

  • SITL-tested for Copter works just fine
  • Tridge sitl-testing for Plane

UTC0057 -

  • New Lighware SF45B rangefinder
    • 50g
    • 300mW
    • 50m
      • Randy has seen 30
    • US$450 (backorder)
  • Same wire protocol as SF40C
  • Drive-by change to proximity library
  • Removing complication on number of sectors
    • Get-horizontal-distances cleaned up
  • Macros need prens
  • Endian-ness issues…
    • RISC-V?
      • WickedShell: RISC-V is little endian
    • We’re a little-endian autopilot….
  • By assuming LE we can save flash….
  • Big-endian lost the war

UTC0107 -

  • Set position for termination
  • If you know the aircraft setup you can perhaps crash the vehicle more reliably?
  • Commit messages need help
  • Tridge can’t think of an aircraft that this would terminate the aircraft better for
  • Putting nose up is deliberate
  • Original values were required by CASA
  • 2:13 PM] WickedShell: My concern is more does this make it more probable for people to be unable to terminate or end up with really weird termination behaviour, vs the existing one will definitely go bad if you could fly the vehicle…
  • Need to stall and spin

UTC0115 -

  • Require armed state to activate failsafes
  • Copter and Plane do failsafes differently
    • Should we make them common?
  • Plane allows failsafes on ground
    • This is how you test them
  • Copter triggers failsafes but doesn’t undertake actions
  • Plane’s jump-to-DLS is a bit of a pain
  • Tridge likes the FS_OPTIONS stuff from Copter
    • Maybe use a bit to indicate if actions should be undertaken
  • No assumption of being disarmed while flying
  • [12:16 PM] WickedShell: if you arm it motors spin… wait yeah what?
  • [12:17 PM] WickedShell: text doesn’t tell me the action worked, mode change does though?
  • Reset mission on arm?
  • New scheme needed
  • People do fly disarmed
    • Watchdog reset
    • RC not present?
      • Want to RC failsafe
  • Tridge does not want to have the vehicle arm to check failsafe
  • Tridge agrees there is a problem this just isn’t the solution

UTC0128 -

  • Nooploop driver
  • Partner is keen on this one
  • Was done as a visual odometry backend
    • Assumes you’ll get a heading
    • Beacons don’t give you that
  • Needs a good review
  • Peter to review
  • Run it through astyle

UTC0132 -

  • Went in completely unreviewed
  • Should we simply revert this driver?
  • PRs should not be going in without review
  • Someone apart from the author should review a change
  • A comment is not a review!
  • many, many things would have been caught on this PR if it had been reviewed
  • Code authors don’t tend to review their own code well
    • Human nature
  • Changed from “ping” to “APM” on the broadcast ID
    • Safety cases?
    • Callsign….
  • Set-and-save notifications/
  • Reviews will be sought in future
  • Parameters for protocol transfer really not a good look
  • Tom will add patches to revert this and create a PR

UTC0203 - Copter update

  • 4.0.5rc1 will go out as beta today
1 Like