Dev Call Jan 12, 2026

CraigElder · January 12, 2026, 10:48pm

CI Workflows

Release Update

Issus & Pull Requests

github.com/ArduPilot/ardupilot

Plane: Quadplane: add thrust loss detection

master ← IamPete1:quadplane_thrust_loss

opened 08:37PM - 21 Dec 25 UTC

IamPete1

+68 -0

This adds thrust loss detection to quadplane. The thresholds are copied from cop…ter. For thrust loss to be detected the following must be true for one second: - Armed, flying and in throttle unlimited spool state - Desired attitude of less than 15 degrees - Demanded throttle of more than 90% or on the upper throttle limit and above 25% - Vehicle descending - Angle error of less than 30 degrees When thrust boost is detected it will stops the mixer constraining the other motor outputs based on the failed motor. This will only really help if you have more than 4 motors since if you have 4 or less you will have already crashed. Copter has a flight option to disable this check, we will likely want the same. Currently this is active all the time, we may want to ignore transitions (especially for tiltrotors) or change the thresholds. Draft PR for now to get some testing. Another option would be to expose the `motors->set_thrust_boost` to scripting so the check could be done with a script and tailored to a specific vehicle, or by using ESC telem for example.

github.com/ArduPilot/ardupilot

AP_NavEKF3: save RAM by removing use of KH array

master ← tpwrules:pr/ekf-no-kh

opened 02:35AM - 27 Dec 25 UTC

tpwrules

+135 -170

Saves 2.25K global RAM on F4 and 4.5K on F7/H7. ~~Note that the no-longer-used t…emporary array is still accessible as it is in a base class, that should probably be refactored in the future.~~ But the array is definitely optimized out. I did not do any performance testing as the mag fusion is the most frequently called and most expensive fusion, more or less, and that was already addressed. But it should give some more performance too, and does save a few hundred bytes of flash as well. Sequel to #31759 . I made sure my boards behaved well on the bench but did not do any flight tests. I also ran `Tools/Replay/check_replay_branch.py` to show that there were no differences in replay. It did show 20-30 errors in the beacon test (I think), but it showed the same thing without this PR applied so I think it is just slightly broken. There were otherwise no errors.

github.com/ArduPilot/ardupilot

AP_AHRS: factor out a _get_quaternion_for_ekf_type

master ← peterbarker:pr/factor-get-quaternion

opened 12:58AM - 01 Jan 26 UTC

peterbarker

+12 -50

This is the difference between the two bodies which are being factored: ``` --…- /tmp/primary 2026-01-01 11:44:16.251345538 +1100 +++ /tmp/secondary 2026-01-01 11:44:03.630086026 +1100 @@ -1,40 +1,49 @@ { - // backends always return in autopilot XYZ frame; rotate result - // according to trim - switch (active_EKF_type()) { + EKFType secondary_ekf_type; + if (!_get_secondary_EKF_type(secondary_ekf_type)) { + return false; + } + + switch (secondary_ekf_type) { + #if AP_AHRS_DCM_ENABLED case EKFType::DCM: + // DCM is secondary if (!dcm.get_quaternion(quat)) { return false; } break; #endif + #if HAL_NAVEKF2_AVAILABLE case EKFType::TWO: + // EKF2 is secondary if (!_ekf2_started) { return false; } EKF2.getQuaternion(quat); break; #endif + #if HAL_NAVEKF3_AVAILABLE case EKFType::THREE: + // EKF3 is secondary if (!_ekf3_started) { return false; } EKF3.getQuaternion(quat); break; #endif + #if AP_AHRS_SIM_ENABLED case EKFType::SIM: - if (!sim.get_quaternion(quat)) { - return false; - } - break; + // SITL is secondary (should never happen) + return false; #endif + #if AP_AHRS_EXTERNAL_ENABLED case EKFType::EXTERNAL: - // we assume the external AHRS isn't trimmed with the autopilot! + // External is secondary return external.get_quaternion(quat); #endif } ``` It's clear that the only substantial difference here is the SITL case which is marked as "can't happen". In the case it does happen then we will just return the SITL results, which is really what we should be doing anyway. This method is only used for logging/telemetry.

github.com/ArduPilot/ardupilot

Step 3: AP_Math: Update shaping to use improved jerk limited algorithm.

master ← lthall:20260101_SqrtJerk_Controller

opened 11:02PM - 03 Jan 26 UTC

lthall

+234 -134

The current real time S-Curves have two problems that this PR addresses: 1. The…re is a small overshoot as it approaches the target position in some combinations of input. 2. The deceleration it generates is significantly less than the limits specify. This PR produces the full deceleration specified by the square root controller and maintains the time constant based approach to the position target. This time constant based approach makes this algorithm act as a filter making it suitable for noisy or inputs and means that the algorithm does not oscillate or buzz near the set-point. Below is a plot showing the original square root controller, the current Real Time S Curve, This Jerk Square Root algorithm and the standard rectangular jerk S Curve (Optimal). There are two reasons we would choose not to use the Optimal solution: 1. Computational Complexity. 2. Lack of any soft approach to set-point. Both of these make Optimal less appealing to high update rate applications. <img width="907" height="1022" alt="image" src="https://github.com/user-attachments/assets/81f10d8b-72c5-4f5e-b9ea-1c55f5295132" /> Meant to be merged after https://github.com/ArduPilot/ardupilot/pull/31870 and https://github.com/ArduPilot/ardupilot/pull/31848

github.com/ArduPilot/ardupilot

AP_Param: work around Clang `-ftrapping-math` issue

master ← tpwrules:pr/clang-trapping-math-fix

opened 09:34PM - 04 Jan 26 UTC

tpwrules

+16 -8

In Clang (independent of OS or architecture), if three factors are true about a …static variable initialization (const or not): * `-ftrapping-math` is turned on * a double literal needs to be rounded and doesn't exactly equal a float * the initialization involves something which isn't constexpr then Clang will decline to do constant initialization, unlike GCC, or Clang if any one of these factors is not true. Constant initialization is where the object is initialized as part of the memory image and is ready to go when the program is loaded. Instead, Clang will generate code at runtime to do the initialization. This is fine in a vacuum, but it makes lots more variables suddenly vulnerable to the static initialization order fiasco. In particular, many of the `AP_Param::Info` tables for the parameter system meet all the factors. The use for these tables, `AP_Param::setup_object_defaults`, is part of constructors of objects which are usually constructed in a different file than the table. Due to the new runtime initialization, there is no longer a guarantee that the table has meaningful data at the time of that setup call. This causes many parameter defaults to not be loaded and the system generally crashes and burns. This PR addresses the issue by fixing the non-constexpr casts in `AP_VAROFFSET` and `AP_CLASSTYPE` to use things permissible in constexpr. The compiler then constant initializes these tables again, parameters get their defaults, and everybody is happy. This is probably easier than fixing all the double literals to be floats. This fix could be "locked in" and be made to cause a compiler error in case of a problem by declaring all tables `constexpr`, but that is a lot of effort. Examination of the binary symbol tables shows that all `var_info` tables are correctly constant initialized now. However, that examination also shows that other things are being initialized differently (unsure if they create an ordering problem): * all the Motor tables in SIM_Frame.cpp * a few constants which use floating point functions like AP_OLC::GRID_SIZE * a whole bunch of "MergedGlobals" symbols that I don't care to figure out Given that the vehicle starts up now instead of crashing, I'm choosing to believe these latter problems are minor and can be fixed later. This is also better than removing the flag and letting random crashes from ignored checks come back. But the ultimate fix may still be something different. The new macros are also probably more defined and safer than the old ones so we should keep this change in regardless. The whole situation is super weird so I also filed an issue with LLVM, but if there's something to fix I expect a very long time before we can remove any workarounds. Tested that there is no compiler output change on CubeOrange, so the new definitions do not trip up GCC. I also ran the Copter test suite on macOS and many things pass so it's no longer completely broken there. Before EKF3 wouldn't even start because its enabled default didn't load, and turning on terrain would instantly crash. It's unclear why it wasn't so bad on Linux Clang, maybe the version or linker behavior is significantly different.

github.com/ArduPilot/ardupilot

Move SITL to using hwdef system

master ← peterbarker:pr/sitl-hwdef

opened 05:59AM - 07 Jan 26 UTC

peterbarker

+371 -210

No compiler output change: ``` Board,AP_Periph sitl_periph_battery_tag,* sit…l_periph_battmon,* sitl_periph_can_to_serial,* sitl_periph_gps,* sitl_periph_universal,* ``` ``` Board,antennatracker,blimp,copter,heli,plane,rover,sub SITL_arm_linux_gnueabihf,*,*,*,*,*,*,* SITL_x86_64_linux_gnu,*,*,*,*,*,*,* sitl,*,*,*,*,*,*,* ``` Still a bunch of work to do in future PRs - defines to move from SITLBoard into hwdefs, moving `defaults_periph.dat` outside of the ChibiOS directory and use it, ...

github.com/ArduPilot/ardupilot

AP_TemperatureSensor: Add additional sensor slots

master ← BristolFlightLab:support-additional-temperature-sensors

opened 10:47AM - 07 Jan 26 UTC

rob-clarke

+84 -0

Add parameters to support up to 15 temperature sensor instances Built code ad…ding `--define AP_TEMPERATURE_SENSOR_MAX_INSTANCES=15` to the configure stage. Tested in SITL using `SIM_Temperature_SHT3x` by setting: - TEMPF_TYPE = 8 - TEMPF_BUS = 1 - TEMPF_ADDR = 68 (0x44) - TEMPF_SRC = 3 - TEMPF_SRC_ID = 1 Observed that battery_temp in Mission Planner status tab was being updated.

github.com/ArduPilot/ardupilot

Provide a method to get a binary log of all serial data

opened 07:46PM - 08 Jan 26 UTC

Ryanf55

DevCallTopic

## Feature request **Is your feature request related to a problem? Please descr…ibe.** I have a serial port, currently configured for a mavlink connection. The autopilot is not accessible when the vehicle is assembled. I want to have an easy way to decode traffic on the port. If it was, I'd use a Saleae logic analyzer, decode as async serial, and then run use pymavlink to decode the stream. I cant use the mavlink inspector in mission planner because this is an internal connection in the vehicle, and some packets on this mavlink interface are not also repeated on the GCS link. **Describe the solution you'd like** Provide an option to enable logging for the serial port, which will log all data to a binary file on the SD card. I can then use mavftp to transfer this to my computer and interpret the contents. **Describe alternatives you've considered** Turning on LOG_REPLAY does not log serial port traffic. **Platform** [ ] All [ ] AntennaTracker [ ] Copter [ ] Plane [ ] Rover [ ] Submarine

github.com/ArduPilot/ardupilot

AP_Airspeed: change primary if current not set to use

master ← IamPete1:airspeed_primary_update

opened 12:47AM - 11 Jan 26 UTC

IamPete1

+147 -42

If the user or the automatic health check sets a sensor to not be used and it is… the primary currently this results in no airspeed sensor being used. This is because health is the only metric used for automatic switching of primary. With this change the primary will switch to a sensor which is both healthy and set to use. Currently to disable use of the primary airspeed the user would have to change the primary param. ~~We may want to make it clearer to the GCS when the primary has switched, but it is available in the MAVLink AIRSPEED message.~~ I have added a send text This also adds a arming check that the primary sensor must be set to use if there is a usable sensor. This also fixes a error in the log documentation for `ARSP.Pri`, it is a index number not a bool for each instance (although alternatively we could change it to a bool). This also adds tests for the new primary switching.

github.com/ArduPilot/ardupilot

Airspeed: Health check: work per sensor instance

master ← IamPete1:airspeed_ekf_check_per_instance

opened 02:52PM - 11 Jan 26 UTC

IamPete1

+25 -12

Currently the airspeed health check using the EKF innovations only does not chec…k if the EKF is using the sensor its disabling. On master you currently get this: <img width="1386" height="884" alt="image" src="https://github.com/user-attachments/assets/a6734817-4434-4ea1-8392-eacdfa92e00e" /> The health probability of both sensors is changed. This means that both sensors are disabled. <img width="283" height="243" alt="image" src="https://github.com/user-attachments/assets/4fb1f127-e5f7-4748-ad4a-b465f69e05e3" /> The innovations recover when the sensor is marked not to use because the EKF is no longer using the bad sensor. So it cycles being disabled and re-enabled. This is the readings of the two sensors, the second sensor is reporting the correct airspeed but is still disabled: <img width="1386" height="456" alt="image" src="https://github.com/user-attachments/assets/69c3e34a-9b09-4039-82f0-d5c256cfcdc1" /> With this patch only the bad sensor is disabled: <img width="1380" height="769" alt="image" src="https://github.com/user-attachments/assets/b2479faa-8b67-4a7e-b831-d4a565bc9af7" /> The second sensor remains healthy. Note that in the disable/re-enable cycles the health only goes to 0 very briefly. The low health triggers the airspeed to not be used and thus there is no EKF data so it starts recovering. <img width="1380" height="769" alt="image" src="https://github.com/user-attachments/assets/7717b8ae-53d2-44eb-94a3-f9747c8e433a" /> Because the current primary airspeed selection logic only checks health this means that the second sensor is never used (CTUN.AsT 3 is EKF synthetic). <img width="1380" height="769" alt="image" src="https://github.com/user-attachments/assets/0b96aea4-17e5-49ce-b0bd-4a2621bf867b" /> With https://github.com/ArduPilot/ardupilot/pull/31889 the primary will change on based on use and the second sensor will be used. This results in more use of airspeed without falling back to synthetic: <img width="1507" height="839" alt="image" src="https://github.com/user-attachments/assets/227de449-6af9-4764-bfa0-999c6a455caa" /> Note that there is still small section of synthetic where the EKF is rejecting airspeed but the airspeed sensor has not yet been disabled.

github.com/ArduPilot/ardupilot

vehicles: use QGroundControl mode names

master ← tpwrules:pr/qgc-modes

opened 11:21PM - 11 Jan 26 UTC

tpwrules

+86 -69

QGroundControl now uses these names directly through MAVLink `AVAILABLE_MODES`. …Update them to be nicer for users. Worth noting that these are also the strings fed in to the speech synth. They are also used in other status messages so those messages will change slightly too. Tested that the modes appear nicely in QGC now. Also had to add the infra to Rover to get full names. Partial fix of #31761 . We need this (or something equivalent) for 4.7 to avoid all the names being capitalized in QGC.

Partner Call Reminder

CraigElder · January 13, 2026, 5:26pm

Attendee count (max): 16

UTC1101 - CI ci · Workflow runs · ArduPilot/ardupilot · GitHub

Just the colcon stuff this week

UTC1104 - vehicle update ArduPilot 4.6 issues list · Issue #28612 · ArduPilot/ardupilot · GitHub

Been very quiet
Need to understand fix ocean tile causing zero altitude for land tiles by tridge · Pull Request #23 · ArduPilot/terraingen · GitHub
- Strategy for rolling out the fix to users
- Quite a number of vehicles may have bad terrain data
- Lost a plane diving into the ground
- Tiles within 100km of the coast
- Maybe use a user-alert?
- Blog post
- Facebook?
- Check altitude delta in our code?
- Use the version header?
- Increase required version
  - Would invalidate all downloaded tiles
  - Or just do extra checking on older versions of the tiles
  - If all data points are zero then that’s also OK
  - Just a mixture of zero and non-zero
- Opportunity to review the terrain data?
- 1097m error in NZ…
- Check against SRTM should get the same data source
- MAVProxy was getting the right heights
  - Didn’t go via the conversion-to-a-dat-file
- We will increase minimum version of tile required for things that have a mixture of zero and non-zero
  - When we load a tile we check the minimum version required
  - Add to that check
  - Treat as invalid if mixture of zero and non-zero
  - Only do the new check when disarmed as we probably still want to use a tile for RTL so we don’t run into terrain
  - Fully-populated non-sparse if you have downloaded
Probably a 4.6.4 patch release required before 4.7.0beta1

UTC1133 - Improve description fields in mavlink messages by peterbarker · Pull Request #459 · ArduPilot/mavlink · GitHub

Clarify mavlink message descriptions
Merged!

UTC1134 - Plane: Quadplane: add thrust loss detection by IamPete1 · Pull Request #31764 · ArduPilot/ardupilot · GitHub

Add thrust loss detection
Constrains on second-highest motor output
Might move you from descending to being able to hover again
Are the Copter threshold valid?
Do we want to do this in C++ at all?
Should we not do this in transitions?
Maybe not enable this by default?
Tilts might have issues as we push the forward propellors up a lot
But we still want this to kick in and save your vehicle even if you are in transition
- In transition includes qassist
50% vs 66% on an x8
User testing showed a difference between which motor was out on a hexa
Partner needed these patches to be able to survive any one motor outage
- Otherwise the tilts were problematic
Needs an option bit

UTC1152 - AP_NavEKF3: save RAM by removing use of KH array by tpwrules · Pull Request #31797 · ArduPilot/ardupilot · GitHub

Save RAM by removing KH array
Merged!

UTC1153 - remove HAL_CHIBIOS_ARCH_FMUV5+ by peterbarker · Pull Request #31831 · ArduPilot/ardupilot · GitHub

Remove hal gmuv5
Merged!

UTC0001 - AP_AHRS: factor out a _get_quaternion_for_ekf_type by peterbarker · Pull Request #31834 · ArduPilot/ardupilot · GitHub

Factor out a method in the AHRS
Merged!

UTC0002 - AP_Notify: stop using ownptr for SH1106 by peterbarker · Pull Request #31837 · ArduPilot/ardupilot · GitHub

Stop using ownptr in sh1106
Merged!

UTC0006 - Step 3: AP_Math: Update shaping to use improved jerk limited algorithm. by lthall · Pull Request #31848 · ArduPilot/ardupilot · GitHub

Update shaping to use improved jerk-limited-algorithm
Leonard thinks this should be able to go in this week…
PH will look, probably on the weekend

UTC0009 - AP_Param: work around Clang -ftrapping-math issue

-fsingle-precision-constant
Merged!

UTC0012 - Move SITL to using hwdef system by peterbarker · Pull Request #31868 · ArduPilot/ardupilot · GitHub

Use hwdef system for SITL
Tridge and Thomas want to keep looking at this

UTC0018 - AP_TemperatureSensor: Add additional sensor slots by rob-clarke · Pull Request #31871 · ArduPilot/ardupilot · GitHub

Add more temperature sensors
Some servo is sending out temperature sensor messages
Merged!

UTC0019 - AP_TemperatureSensor: Add additional sensor slots by rob-clarke · Pull Request #31871 · ArduPilot/ardupilot · GitHub

More temperature sensors
Merged!

UTC0028 - Provide a method to get a binary log of all serial data · Issue #31880 · ArduPilot/ardupilot · GitHub

Get binary log for a serial log
Peter will push ahead with the bytebuffer thing
See Ryan’s mavlink decoder for saleae, too!

UTC0030 - AP_Airspeed: change primary if current not set to use by IamPete1 · Pull Request #31889 · ArduPilot/ardupilot · GitHub

Change airspeed sensor if primary use becomes zero
You get told when you change too
Merged!

UTC0035 - Airspeed: Health check: work per sensor instance by IamPete1 · Pull Request #31893 · ArduPilot/ardupilot · GitHub

Sets sensor health on a per-sensor basis
EKF will only disable one of your sensors, not both of them at once
You still end up oscillating
We should calculate the innovation of an unused air speed sensor so we don’t start to use it unless the innovation is low
You can set a bit to stop using the airspeed sensor for the flight
- Primary switching could be improved?
Would need to fuse with and without the airspeed sensor to properly work out whether we should use it!
Can be merged once CI has passed

UTC0047 - vehicles: use QGroundControl mode names by tpwrules · Pull Request #31897 · ArduPilot/ardupilot · GitHub

Revise human-readable strings
Qloiter is pronounced cloister in QGC
Gcs_name function?
Thomas will keep poking

UTC0054 - modules/DroneCAN: update to fix new Python compatibility issues by tpwrules · Pull Request #31898 · ArduPilot/ardupilot · GitHub

Update dronecan submodule to bring in some Python fixes
Merged!

Dev Call Jan 12, 2026

Attendee count (max): 16

UTC1101 - CI ci · Workflow runs · ArduPilot/ardupilot · GitHub

UTC1104 - vehicle update ArduPilot 4.6 issues list · Issue #28612 · ArduPilot/ardupilot · GitHub

UTC1133 - Improve description fields in mavlink messages by peterbarker · Pull Request #459 · ArduPilot/mavlink · GitHub

UTC1134 - Plane: Quadplane: add thrust loss detection by IamPete1 · Pull Request #31764 · ArduPilot/ardupilot · GitHub

UTC1152 - AP_NavEKF3: save RAM by removing use of KH array by tpwrules · Pull Request #31797 · ArduPilot/ardupilot · GitHub

UTC1153 - remove HAL_CHIBIOS_ARCH_FMUV5+ by peterbarker · Pull Request #31831 · ArduPilot/ardupilot · GitHub

UTC0001 - AP_AHRS: factor out a _get_quaternion_for_ekf_type by peterbarker · Pull Request #31834 · ArduPilot/ardupilot · GitHub

UTC0002 - AP_Notify: stop using ownptr for SH1106 by peterbarker · Pull Request #31837 · ArduPilot/ardupilot · GitHub

UTC0006 - Step 3: AP_Math: Update shaping to use improved jerk limited algorithm. by lthall · Pull Request #31848 · ArduPilot/ardupilot · GitHub

UTC0009 - AP_Param: work around Clang -ftrapping-math issue

UTC0012 - Move SITL to using hwdef system by peterbarker · Pull Request #31868 · ArduPilot/ardupilot · GitHub

UTC0018 - AP_TemperatureSensor: Add additional sensor slots by rob-clarke · Pull Request #31871 · ArduPilot/ardupilot · GitHub

UTC0019 - AP_TemperatureSensor: Add additional sensor slots by rob-clarke · Pull Request #31871 · ArduPilot/ardupilot · GitHub

UTC0028 - Provide a method to get a binary log of all serial data · Issue #31880 · ArduPilot/ardupilot · GitHub

UTC0030 - AP_Airspeed: change primary if current not set to use by IamPete1 · Pull Request #31889 · ArduPilot/ardupilot · GitHub

UTC0035 - Airspeed: Health check: work per sensor instance by IamPete1 · Pull Request #31893 · ArduPilot/ardupilot · GitHub

UTC0047 - vehicles: use QGroundControl mode names by tpwrules · Pull Request #31897 · ArduPilot/ardupilot · GitHub

UTC0054 - modules/DroneCAN: update to fix new Python compatibility issues by tpwrules · Pull Request #31898 · ArduPilot/ardupilot · GitHub

UTC0055 - partner call tomorrow

UTC0057 - close