Dev Call April 20, 2026

Development Team
2

Attendee count (max): 17

UTC2301 - CI

  • Blimp test
  • Slowdown seems to be their infrastructure
  • Release took a long time
  • 30-odd PRs to fix bogged things down
    • AI-created
    • Require approval for first-time contributors
    • Currently “to github”
    • Maybe change to “to ardupilot”
      • Yes, done

UTC2310 - release update

  • Custom message crc error thing

  • Need to revisiting the MAVn_DEVID thing

    • It’s still a source of major confusion

  • Mavlink logging docs need improving

  • Terrain download stuff that Tim reported

    • Visible as terrain pending bouncing up and down
      • Fix the TERR message to log the number of requests made to the GCS
    • Thomas looked at it
    • Logs show there was no attempt to grab terrain from GCS
    • Cache being thrashed
    • Default cache size was 12
      • 12 costs 12*1800 bytes
        • 21kB
      • Reasonable for 15m/s with 100m spacing
      • Tim’s using 30m spacing and moving faster
      • With lookahead in fixed wing it is constantly trying to pull in more terrain tiles
      • Reproducible in SITL
      • Lower the terrain cache size TERR_CACHE_SZ
      • Use maybe 50
        • Yay H7
    • When circling there’s a tangent vector scanning the landscape!
      • Curved lookahead?!
    • Constant re-evaluation of terrain is a good thing
    • 30m or less spacing on Copter / with terrain-based navigation could be a real problem
      • Terrain failsafe may trigger
      • RTL can also use terrain….
        • Would return at home alt interpreted as above home if that fails

  • Elrs

    • Joystick and regular RC
    • RC off all the time
    • Flying just with joystick
    • When telemetry is broken
    • Switches back to the regular RC which has never been used

  • Buffer overflow in CAN logging is critical

UTC2347 - common.xml: update MAV_CMD_DO_SET_MISSION_CURRENT desc by peterbarker · Pull Request #498 · ArduPilot/mavlink · GitHub

  • Update XML
  • Merged
  • Some confusion on the wording

UTC2357 - https://github.com/ArduPilot/ardupilot/pull/32416

  • Add PilotGaea board
  • Merged

UTC2359 - https://github.com/ArduPilot/ardupilot/pull/32605

  • Fix build problems with test_build_options.py
  • Merged!

UTC0001 - https://github.com/ArduPilot/ardupilot/pull/32634

  • Fix payload place and land altitude bug
  • Issues found

UTC0001 - Convert QURT HAL to using hwdef system by peterbarker · Pull Request #32728 · ArduPilot/ardupilot · GitHub

  • Move QURT over to using hwdef system
  • Actually uses two toolchains
    • Two architectures….
  • Merged!

UTC0006 - Plane: add input shaping to roll and pitch controllers by IamPete1 · Pull Request #32743 · ArduPilot/ardupilot · GitHub

  • Add input shaping to roll and pitch controllers
  • Real-time scurves used to limit control output
  • Need to understand step changes in estimator output
  • Need to see what happens with upset recovery
    • We don’t limit ourselves to roll rates when doing upset recovery
  • Ardupilot is a submodule of web tools now
    • Compiles to webasm
  • Should be able to use the parameter revert script

UTC0020 - AP_Compass: report specific failure reason when fit rejected by christianpetri · Pull Request #32757 · ArduPilot/ardupilot · GitHub

UTC0040 - Issue with throttle mix and quadplane hard landings · Issue #32759 · ArduPilot/ardupilot · GitHub

  • Throttle adjustment for VTOL land
  • Abort and land use the same numbers
  • Don’t force mix-max during VTOL landing descent
  • Learning hover throttle doesn’t really happen on a quadplane
    • Doesn’t spend long enough at hover throttle in quadplane
    • And then you find you were in a wind when this happened
    • Learnt very low hover throttles are very bad
  • Lua script to adjust the hover throttle for massive payload drops
  • We also have learn-but-don’t-save option
  • Descending fast in a quadplane leads to instability
    • So allocating output to stability and increasing your descent rate can be counterproductive!
  • #32759
    • Difference between MOTB.ThrAvMax and QTUN.TMix
    • Former was really high
    • The latter was 0.5
    • In some part of the code the tmix better reflects what is being used
    • Output-arm-stabilizing uses one but might want the other?!
  • Leonard will have a look
  • There is a web tool to help you understand these numbers
  • Probably a 4.7 candidate for the mix/max change

UTC0057 - hwdef: Add JFB200 of JAE flight controllers by n-ito2222 · Pull Request #32786 · ArduPilot/ardupilot · GitHub

JFB2000 board

  • Reviewed it
  • We agreed that we would have provisional allocations on board IDs

UTC0111 - AR_Motors: Make power limiting eaiser for end-users by stephendade · Pull Request #32788 · ArduPilot/ardupilot · GitHub

  • Power limiting for Rovers
  • Stop motors overheating due to over-watt
  • When testing on a real Rover the low-passed power added lots of lag
    • Could take 15 or 30 seconds to hit the power limit
    • Made it difficult to understand when the limiter was in effect
  • This PR lowers the filter time constant
  • Maybe not on the statustext?
    • Is hitting your limits an error?
    • This acts in manual mode!
    • Disconcerting on the throttle
    • Want to ensure the user knows what is happening
  • Stephen thinks that the watt limiter being active is always a sign of a bad thing happening
    • Thomas thinks he wouldn’t want to be nagged!
    • MOT_THR_MAX to 80% each?
      • Doesn’t work with DC motors?
  • Statustext will be removed…
  • Stephen will come back with this one

UTC0122 - Surface tracking: preserve full pilot climb/descent authority by lthall · Pull Request #32789 · ArduPilot/ardupilot · GitHub

  • Preserve full pilot climb/descent authority
  • Surface tracking can saturate
  • Allows for max climb even in the face of terrain
  • Not without problems at the bottom of a hill you might end up with a climb rate above your maximum but it is better than the double you could get at the moment

UTC0127 - Add support for Shanghai Fudan Microelectronics (FMSH) FM25Q series NOR Flash by gengchi-fuan · Pull Request #32796 · ArduPilot/ardupilot · GitHub

  • Add support for FM25Q autopilot
  • Already lots using this?!

UTC0131 - Augment sim_vehicle.py with a GUI · Issue #32802 · ArduPilot/ardupilot · GitHub

  • Request for a GUI for SITL
    • Scenario-based seems OK
  • Aside:
    • MissionPlanner + quadplane tilt
    • Very difficult to get this working in MissionPlanner SITL
    • Issue was opened
  • If you have ideas on what should be in this “starter gui” please add to issue
  • mavagent?!

UTC0141 - autotest: add test for plane-wind-failsafe.lua by peterbarker · Pull Request #32807 · ArduPilot/ardupilot · GitHub

  • New autotest for Plane
  • Discussion around Lua debugging
  • [11:43 AM]NateMailhot: Recursive self repairing script?
  • Peter has coverage working

UTC0144 - Tools: break callisto SITL config to a common method by Ryanf55 · Pull Request #32828 · ArduPilot/ardupilot · GitHub

  • Break Callisto starting method out into a method
  • Do we want to use more Python features

UTC0149 - beta

  • Will be trying to keep to the 2 week cycle
  • Tag your PRs!

UTC0155 - discussion on votes

  • Should ask user before raising vote
  • Follow the instructions (dev-team user guide!)

UTC0156 - looking at log

  • Throttle stuck at 50%
  • Everything is just oscillating
  • roll/pitch commands pulling the average up throttle
  • During descent it builds the vertical speed
    • Natural instability bites
  • If the mix was at 0.1 would the vehicle have done better?
  • Descent rate was commanded at 1.5 but got 6
    • Instability let the speed build up
  • Leonard thinks there might be more to this
    • Throttle was pegged
    • He thinks this would do this regardless of what you did
    • Previous experience shows that loiter stabilised the aircraft
      • Why does this work?
      • Maybe there’s an impulse here somewhere?
        • A momentary reduction in throttle?
    • Maybe if we hit a stability limit then stop trying to descend
      • Or if you exceed a descent rate then command zero
    • Leonard thinks tune the aircraft to handle the maximum descent rate
    • Might need to limit the descent rate more on the aircraft
    • Turbulence etc etc….
  • So too-rapid and the vehicle may not recover
  • But sometimes too much stabilisation required and the quadplane can’t descend

UTC0213 - close