Dev Call May 30, 2022

Issues & Pull Requests

Plane

Copter

GSOC

1 Like

Attendee count (max): 16

UTC1102 - AP_Param: try both native and AP::FS filesystems for posix systems by Williangalvani · Pull Request #20193 · ArduPilot/ardupilot · GitHub

  • Fixes Navigator board
  • Sub wants a backport
  • Been in for master months now
  • Tridge nad Randy will discuss when to backport this

UTC1104 - Clarify get_pilot_desired_lean_angles and fix limit by lthall · Pull Request #20428 · ArduPilot/ardupilot · GitHub

  • “Clarifies” get pilot lean angles
    • Fixes shape at extremities on the diaganonals
  • Now a common library function so quadplane gets same fixes
  • Not a critical fix
  • merged!

UTC1107 - AP_Common: Location: move offset method implem. to header by peterbarker · Pull Request #20846 · ArduPilot/ardupilot · GitHub

  • Move implementation of a method to header
    • Loses lots of bytes on MatekF405
  • [9:09 AM] Peter Hall: I have seen odd stuff like this before.
  • [9:10 AM] Peter Hall: I usualy blame “alignment” or something else equally vague.

UTC1111 - AP_Periph: added RNGFND_MAX_RATE parameter by tridge · Pull Request #20861 · ArduPilot/ardupilot · GitHub

  • RNGFND_MAX_RATE parameter to limit rate at which rangefinder messages are sent on AP_Periph
  • 100Hz rangefinder you should be able to look through trees….
  • Do aggregation on the rangefinder?
    • So send longest or shortest or average or what from the rangefinder?
      • This punts that and the autopilot can do that aggregation
  • 2 bytes of rangefinder is 24 bytes on the DronCAN bus…
  • Flight-tested on a benewake TF-03

UTC1115 - Replay: fixed replaying with EK3 source set changes by tridge · Pull Request #20862 · ArduPilot/ardupilot · GitHub

  • Fix replay with EK3 source changes
  • Merged!

UTC1117 - AP_Baro: auto-set field elevation from origin by tridge · Pull Request #20867 · ArduPilot/ardupilot · GitHub

  • Auto-set field elevation from origin in baro
  • Matches EKF origin
  • Aviation altitude more correct
  • Should get Ryan to look

UTC1124 - AP_Filesystem: fixed gets() posix compat function by tridge · Pull Request #20868 · ArduPilot/ardupilot · GitHub

  • Fixes bad fgets implementation
  • Merged

UTC1127 - EKF: fixed ring buffer implementation by tridge · Pull Request #20870 · ArduPilot/ardupilot · GitHub

  • Fix for the EK3 time horizon buffers
  • Currently if the buffer is empty and you recall that element it fails
  • Only works if there’s >1 element in the buffer
  • No test suite
  • Rewritten….
  • New unit tests written
  • Test suite needs looking at and ensuring it is correct
  • Other issues:
  • Peter will review
  • Others really should review
  • Symptoms of old behavior:
    • Unsure
    • Only get half your data?
  • EKF IMU buffer also needs looking at!
    • Would it hit all lanes at once?
  • Do we really need >1 time horizon buffer, one for each core?
  • Saves a lot of flash

UTC1143 - Copter: Auto mode gets attitude control by rmackay9 · Pull Request #20851 · ArduPilot/ardupilot · GitHub

  • merged before call

  • Lean forward 20 degrees for 5 seconds

  • ATTITUDE_TIME

  • ATTITUDE_RATE_TIME?!

    • Auto is “unsupervised”
      • Leonard reckons anything we do in auto would be “safe”, i.e. not going to crash the aircraft
        • This current PR is essentially alt-hold
          • Won’t reduce altitude

UTC1143 - Plane: stable 4.2.x issues list · Issue #15941 · ArduPilot/ardupilot · GitHub

  • Plane 4.2.1
  • 4.1 → 4.2
    • Q_M_PWM_MIN and MAX both being zero?
    • Via param load?
    • D-short won’t initialise
    • Sometimes only 1 set?!
    • Maybe people are auto-correcting it back to what it was before upgrade!?
      • I.e. doing a param-diff and trying to “fix” things
    • Copter has a check for it
      • Tridge wants to fix automatically
    • Tridge will try to reproduce
  • Quadplane i-gain/p-gain ratio needs to be checked
    • Many people setting i>p which is very bad
    • A specific arming check is coming in to catch this. Igain should always be less than your pgain
    • In feed-foward i should be feed-forward
      • Otherwise i<p!
      • Gyrocopter needing really big i-gain…
        • That was a feed-forward thing
    • Some planes will definitely flip
      • PH thinks “not all planes”
      • Any COG change will be problematic
        • Oscillating on I…
    • Maybe an option bit to allow i>p?
  • Tridge thinks we really do need to be checking parameters for the user
    • QP XY gains
      • Must keep these in proportion!
        • People don’t
      • We should check them
      • Tridge want sto add a pos_xy_mul
        • Defaulting to 1
      • Leonard not in favor
      • This isn’t even documented!
      • Rate gains in MissionPlanner are done with a single slider
        • 10-year-old screen
          • Nobody knows what it does….
    • Tridge thinks we need to change the way we think about parameters
      • Be more careful as to what we allow
    • Tridge wants a case before allowing i>p
      • Copter with big CG offset
        • Big P was problematic
        • Bendy vehicle
    • Fix on GCS?
      • Too slow to get it out to embedded QGC on hardware
    • New parameter markup, “@Ratio”?
      • To compare against other parameters
    • This would break a whole bunch of things?
      • They fix their parameters
      • We did this with Q_ASSIST_SPEED
        • People weren’t setting it
        • We made people change it to -1
  • Leonard wants flexibility
    • Prearm check skirts the line
    • Tridge would add an option bit to allow i>p
    • Larger the multicopter i-to-p smaller i term
      • As you go smaller you tend to have higher i term
        • Potential benefits having i>p
    • Has been bitten by not allow people to do things, to go into that corner
    • I not related to p on Copters…
    • Separate prearm bit for PIDs?
      • The “I’m an expert” prearm bit?
  • Current proxy rations are from a discussion with Paul from a long time ago
    • No methodical analysis on them
    • You might not have to reduce position-p as much as velocity-p
    • Single knob might not do it
  • Leonard might be OK with the MUL thing… makes him throw up a bit in his mouth, but might be able to look away….
  • We have a lot of configurability
    • Not just stupid people making these mistakes…
    • Leonard deliberately using colorful language
      • People not taking the time and being irresponsible
        • “This is why we see so many aircraft flying like shit and costing lots of money….”
  • Need a section on the Wiki for the position controller
    • Most Leonard tends to do is divide by 2
    • Rarely happens
    • Defaults tend to work really well
      • Lean agle is acceleration and the rules get very simple
      • Very large or very unhealthy multis sometimes need reduction
        • Carry-on from issues in the attitude controller
        • Unless it is over 100kg you generally don’t need to reduce
        • Quadplanes are different again, ‘though
    • If you’re having xy problems your problems are probably further up

UTC0026 - Copter update

  • Copter/Rover: 4.2.0 issues list · Issue #20192 · ArduPilot/ardupilot · GitHub

  • 4.2.1rc1 is out in Beta testing

  • Maybe some d-shot fixes?

    • But still some reported issues
  • New issues, only one that’s important

    • Lidarlite broken?
      • Only on H7?
      • 4.1.5 broke too?
        • I2c rangefinders in general?
      • Callistos are running i2c and they’re working just fine
      • Lidarlite has special i2c handling
        • Always been problematic
          • Set_split_transfers call
          • Fpga on their side implements i2c
            • Within the i2c spec, they say
              • Every implementor to use the pulsedlight struggled
                • Benewake is much easier to work with as a company
  • Missions may complete while vehicle is disarmed?

    • Plane has special code to not complete some items e.g. takeoffs
  • Full double-notch?

    • Depends on Plane….

UTC0040 - Tracker update

UTC0045 - GSoC update

  • ROS1 rather than ROS2
  • Want to do ROS2 but mavros isn’t there yet
  • Cartographer etc etc for ROS
  • Sanket and Shiv are all over things
  • PH’s student not so much ATM
  • Half our students engaged….

UTC0047 - partner’s meeting tomorrow

UTC0047 - close

I just had a thought about the arming checks discussion.
How about a Lua script?
It could include a standard set of parameter value checks, with maybe a flag for each check as to whether it’s a warning or error.
The nice thing about this is that partners (or anyone really) can choose not to use it, or have their own version with customized checks added or removed. It would also not be GCS dependent, errors/warnings could be sent as standard GCS text messages.

1 Like

I had a go at implementing this. This is a DRAFT PR for comment/suggestions.

1 Like