Dev Call Aug 17, 2026

Development Team
1

CI Workflows

Release Update

Issues & Pull Requests

2

Attendee count (max): 14

UTC2258 - before-agenda

UTC2302 - CI ci · Workflow runs · ArduPilot/ardupilot · GitHub

  • CANGPSCopterMission really needs looking at (still)

  • Maybe Buzz’s shared memory model might be nice?

  • Tridge has patches for moving to unix domain sockets

UTC2305 - Release update

  • Holybro AM32 ESCs producing errors?

    • Probably a single user’s setup

    • Bdshot

    • Original spec didn’t specify polarity

    • Two different platforms used two different mechanisms

    • Reversing?

      • Combine rcmask with 3dmask

  • QMC5883P not compiled in by default is becoming a problem

    • Several users having issues

    • We probably need a one-in, one-out policy

  • We have a gps-for-yaw issue in the EKF

    • Too heavy a reliance on compass

  • Bmc581

    • Ball is currently in partner’s court

  • Method on nullptr

    • Tridge would be happy to merge the PR now

    • Community dev work is great

    • We need a minimal fix for backport

    • Tridge needs to look at the robot’s output

  • Long-standing thrust-vectoring tilt rotors roll axis is backwards

  • Broken for years

UTC2342 - Copter: parameterize rotation rate for flip mode by bnsgeyer · Pull Request #23578 · ArduPilot/ardupilot · GitHub

  • Bill adding rotation rate

  • Rotation rate was 400deg/s

  • Timeout of 2.5s

  • Less than a second to do rotation

  • What multiple *should* we use?

    • 2x seems right

UTC2345 - Plane: Quadplane: enable weathervane in VTOL auto by IamPete1 · Pull Request #33836 · ArduPilot/ardupilot · GitHub

  • Enable weathervane in VTOL auto

  • The robot wants the reset fixed

  • Robot thinks pilot input for yaw is important

    • Autoland and auto-takeoff

    • Probably should follow stick-mixing

  • PH needs to look at the robot output

UTC2351 - CIRCLE (LOITER_TURNS) mission item using S-curves by lthall · Pull Request #33837 · ArduPilot/ardupilot · GitHub

  • SCurves for loiter-turns

  • Circle mode and this are cross-contaminated

  • Leonard would like to use WPNAV_SPEED rather than an angular rate

    • Want to change circle mode likewise

  • So all your waypoints have the same speed!

  • Pirouette would still use the angular rate

  • Leonard’s never look at circle mode

  • At 0 radius in a mission item you use an angular rate

  • Michelle and Peter flew it, was nice

  • Will go in with Peter and Randy’s help

UTC2359 - AP_Follow: discard target state when FOLL_SYSID changes by lthall · Pull Request #33891 · ArduPilot/ardupilot · GitHub

  • Discard target state when FOL_SYSID changes

  • Currently retains old state

  • Doesn’t fix the jitter problem

    • Multi-path jitter stuff is something needs fixing

  • Nobody minds this one

  • So long as someone approves it it can go in

UTC0009 - AP_NavEKF3: correct moving baseline GPS yaw for vehicle attitude by lthall · Pull Request #33706 · ArduPilot/ardupilot · GitHub

  • Correct moving baseline GPS yaw for vehicle attitude

  • More eyes on would be good

  • Looking pretty good!

  • Need to look at the delay thing that the AI brought up

UTC0013 - Scripted Bendy Ruler based Object Avoidance for Plane by timtuxworth · Pull Request #33955 · ArduPilot/ardupilot · GitHub

  • Bendy ruler algorithm script for Plane

  • AP_OAScripting for objects to avoid

  • More stuff to go into the Lua script

    • Different countries, different rules

  • 9kB on Plane

    • OA library

    • >2MB only

    • If it was 500b or 1kB we might have it in by default

  • LocationDB?

    • Is it fair to hold this up for that to go in?

UTC0031 - Filter: HarmonicNotch: fix Quintuple Notch by IamPete1 · Pull Request #33974 · ArduPilot/ardupilot · GitHub

  • Quintuple notch

  • Makes them work

  • Constrains so we don’t overwrite memory

    • “Composite notches” case

  • Nobody tested it properly

  • If you get this patch you might suddenly get more notches

    • In which case you will need to turn it back down

  • We will cherry-pick this back

UTC0042 - Cope with new ListDirectoryWithTime FTP command, fix bugs by peterbarker · Pull Request #1724 · ArduPilot/MAVProxy · GitHub

  • Fix some ftp bugs

  • Merged!

UTC0051 - AP_HAL_ChibiOS: add Agam MegH7 board support by abhijithcv · Pull Request #33978 · ArduPilot/ardupilot · GitHub

  • Add support for Agam MegH7 board

  • Merged!

UTC0053 - AP_InertialSensor: examples: add example to calculate transfer function of EKF input by IamPete1 · Pull Request #33979 · ArduPilot/ardupilot · GitHub

  • Add example to calculate transfer function of EKF input

  • Filter what we’re sending to the EKF?

  • “The angle really is changing, that’s why we have the filter in the pid loop”

    • This is why we lpf in the pid loops

    • Moving average filter is currently what goes into EKF has nulls at the aliasing frequency which is nice

      • The current design is *supposed* to take this unfiltered data!

  • Lever-arm effect?

    • E.g. GPS a long way from CoC

    • We do end up integrating a bunch of noise because of the way the velocity vector is transformed

    • E.g large quadplanes with GPS out on wing

  • PH has seen vehicles where you can see the motor noise in the angular estimates

  • LH considers that to be a “normal” thing

  • Would be interesting to run some experiments

  • Tridge read a paper where it was claimed running filtered data into the EKF would make sense

  • Phase lag being the big problem

  • 50Hz vs 400Hz loop rates

    • We feed significantly different data into the EKF!

    • See pictures

  • Might we end up with PID aliasing?

  • On Copter @400Hz

    • Delayed time horizon

    • Propagate forward using gyro and accel data to come to current time

    • Attitude estimate updates at 400Hz (most states)

    • Other estimates (covariance matrix etc) is done at 100Hz (AKA fusion step)

    • Relies on gyro+accel to do short-term update

    • Quat is updated at 400Hz from the noisy source

  • Ins fast rate loop interactions

  • Will only have a positive impact on people doing everything right

  • Into EKF or just into the “projection” step?

    • Might be a no-downside improvement if we just do the projection with the filtered data

UTC0109 - AP_InertialSensor: assign sensor health once per cycle by andyp1per · Pull Request #33989 · ArduPilot/ardupilot · GitHub

  • Robot says there’s a bug

  • General agreement with the premise

UTC0115 - AP_DDS: fix frame conversion in external odom by Rei23112 · Pull Request #33992 · ArduPilot/ardupilot · GitHub

  • AP_DDS frame conversion fix

  • Marked by Ryan as NeedsTesting

UTC0118 - Ensure Copter always starts at startup location for each test by peterbarker · Pull Request #34003 · ArduPilot/ardupilot · GitHub

  • Assert autotest Copter always start from correct position

  • Merged!

UTC0119 - Sitl shared memory - eg swarms, following, master-slave, virtual peripheral/s, etc by davidbuzz · Pull Request #34019 · ArduPilot/ardupilot · GitHub

  • Anything that has more than one SITL instance can benefit from this

    • Periph doesn’t use this yet

    • In-scope for peripherals

  • Needs to be tested under cygwin

  • Makes stuff *much* faster for Tim

  • Need to make sure RealFlight works

  • Buzz

UTC0134 - Stop state leaking out of various tests by peterbarker · Pull Request #34023 · ArduPilot/ardupilot · GitHub

  • Stops state leaking in the suite

  • Can be merged after fixing the bug

UTC0137 - AP_ADSB: Rename ADSB_TYPE = 1 to be MAVLink by timtuxworth · Pull Request #34032 · ArduPilot/ardupilot · GitHub

  • AP_ADSB: add MAVLink source type to ingest ADSB_VEHICLE

  • Renames the existing thing

  • Slight tooling change but then can be merged

UTC0042 - Copter: do not treat a stale baro sample as evidence of not falling by peterbarker · Pull Request #34033 · ArduPilot/ardupilot · GitHub

  • Parachute deploy fix

  • Roughly 3% chance of parachute not deploying

  • Blocked on scope creep to use a different altitude source

UTC0047 - AP_HAL_Linux: fix CAN regression by landswellsong · Pull Request #34062 · ArduPilot/ardupilot · GitHub

  • CAN regression fix

  • Merged!

UTC0048 - AP_GPS: do not treat a backwards timestamp as a delayed frame by peterbarker · Pull Request #34069 · ArduPilot/ardupilot · GitHub

  • Avoid jitter correction causing GPS frames to appear a minute old

  • Merged!

UTC0050 - AP_Logger: record RADIO_STATUS source by landswellsong · Pull Request #34073 · ArduPilot/ardupilot · GitHub

  • Radio status log source system or source component

  • Can be merged once the ss is a # and the commit structure fixed

UTC0053 - AP_HAL_ChibiOS: add reboot to USB mass storage mode by tridge · Pull Request #34087 · ArduPilot/ardupilot · GitHub

  • Reboot to USB mass storage mode

  • Mavlink and chibios changes

  • Maybe change formatting code to format as exfat?

  • USB interface is the bottleneck

  • Tridge is hoping for 950kB

  • There are boards out there that can do high-speed USB

    • On H7 if you use high-speed you can’t do dual-CAN

  • Tridge doesn’t want to do this at runtime

    • Too much race

UTC0109 - Tools: select matching board for upload on linux by tridge · Pull Request #34088 · ArduPilot/ardupilot · GitHub

  • reboot correct flight controller for apj

  • Merged!

UTC0109 - AP_HAL_ChibiOS: add full-memory crash dumps to microSD by tridge · Pull Request #34094 · ArduPilot/ardupilot · GitHub

  • Add full memory crash dumps to microsd

  • Thomas would like to chat with tridge on this

UTC0119 - close

3

I’m interested to learn more about this. As far as I know, GPS modules run their own filter first right? And it also seems the BMI270 and others do some filtering too…