Issues & Pull Requests
Plane
Copter & Rover
GSOC
Attendee count (max): 16
UTC1100 - compass per-instance parameters
- Failing CI
- AP_Periph GPS is too big on old compiler
- One compass costs more flash, more than one saves flash
- New compiler is fine
- Nobody sees why we should keep the older compiler in CI
- PH will put together a PR to disable it
- DEV_PRINTF in place of console->printf
- Flow control in sys/uarts.txt would be good
- Tridge doesn’t want them to compile them out
- Only see these on error usually
- Tridge doesn’t want to get rid of them globally
- DEV_PRINTF macro is harmless if it produces same result
- Removing them needs case-by-case basis
- Don’t want to worry people too much over trivialities
- Someone is going to have to do a hundred at a time
- This is bulk flash savings - 4kB at least
- Doing the maco change is a reasonable start
- Use different spoolup/spooldown times on quadplanes
- So avoid desync on startup but move from idle to stop fast
- Needs to check which way we’re going when going to idle
- Little more work but it’s close
- Dcm normalization comment added to say the 10 is probably wrong but we need evidence to fix it
- Bootloader update
- Merged
- Remove reference to BRD_PWM_COUNT
- Too many commits
- Two developers already tried to fix it for the new dev but can’t push to the branch
- Harmonic notch filter fixes.
- INS notch turned into harmonic notch but without full capabilities of harmonic notch
- Bit of a hack
- Missed it when merging
- Tridge created a proper parameter conversion
- Fixed notch would be broken when moving to 4.2 from 4.1
- This makes it two harmonic notch filters
- Maybe not a candidate for 4.2.0?
- But 4.2.1
- Different KV motors front and back? This can support that
- Vtol and petrol motors - can have a harmonic notch on both
- Large but relatively mechanical change
- Values of hidden parameters were being used
- I.e. feature disabled but using those params anyway
- But values can affect behaviour of the code
- Saves 228 bytes on Plane on Matekf405
- Mainly removes indirections
- Parameter conversion code will go into 4.2.0
- merged
- Add more keywords to describe a vehicle
- Allows specification of moments of inertia
- merged
- Trim incorrect for tailsitters without pitch offset
- Rotation + trim…
- Get_rotation_and_trim?
- Move Copter filtering up so it can be used in Plane
- Tailsitter landing doesn’t work so well
- How does this interact with Plane’s existing RangeFinder filtering?
- Hasn’t been looked at
- Tridge has a rangefinder complementary filter he should get ready
- Wrapper class with params would be good
- Need to make sure dropouts are handled correctly
- File descriptor zero being closed inadvertently
- Merged
- Plane: add Q_OPTION for RTL always on VTOL RC failsafe for ship landing
- If you get an RC failsafe in vtol mode it will go around an do ship landing appropriately
- Comments need fixing
UTC0003 - Plane: remove support for alt wait by IamPete1 · Pull Request #20548 · ArduPilot/ardupilot · GitHub
- Actively being used
- Alt-wait
- Needs the one user on the planet to actually move off using it
- Add option for LUA script to take on responsibility for a mission item?
- Would still cost us flash to pack/unpack into storage
- This is for dropping gliders from a very long way up
- Have to have the LUA script before we can remove this
- Fixes two issues
- Upset EKF
- Reset ESCs
- Peter will look at aftercall
- Tests for Rover horizontal rangefinder
- And change that to use distance in m rather than cm
- Merged
UTC0027 - Plane: stable 4.2.0 issues list · Issue #15941 · ArduPilot/ardupilot · GitHub
- Plane updates
- Found problem with zero-watchdog in omnibusf4pro
- Probably an issue in d-shot
- May affect 4.1 also
- QRTL error
- Stop flying away in avoidance recover
- Tridge needs to look at
- A few small things for 4.2
UTC0029 - Copter/Rover: 4.2.0 issues list · Issue #20192 · ArduPilot/ardupilot · GitHub
- BRD_PWM_COUNT
- DO_SET_SERVO stops working?!
- The servo x function default is still zero which breaks things now
- Users have to set to -1
- Should they get an arming failure if they don’t set things correctly?
- Lots of users are having issues with it
- Default setup on CubeOrange used to specify pin as two functions
- But BRD_PWM_COUNT kinda made things work
- Can’t default it or we might disable people’s parachute release
- Message length in mission planner means have to keep the message short
- Just document our way out of it?
- Might be able to do better on the RPM pin
- Still some nasty stuff on some Linux boards in terms of defaults
- Randy will look at it
- Still lost memory between 4.1 and 4.2 by ~4kB
- Baseline memory usage increase
- Probably not a single patch
- Second CAN driver uses twice the memory?
- Depends on how you allocate things
-
OSD: hide parameters correctly on var_info2 by tridge · Pull Request #18533 · ArduPilot/ardupilot · GitHub
- Parameter hide issues
- Compass learning on Copter basically worked
- Might be taking a little longer
- Most people don’t use off-diags and scale factors
- Magfit is definitely the best way of doing things
- COMPASS_LEARN=3 should be good on 4.2 for Copters
- Most people don’t use off-diags and scale factors
- Might be taking a little longer
- Dance needed for rotations
- Compass_lean should be able to get orientation
- GSF takes about 3 seconds to determine yaw
- Takeoffs where vehicle was above plate in ground causing temporary but significant incorrect heading at takeoff
- GSF wasn’t fast enough…
- All-but-one-agreeing sometimes happens
- Limit lean angle based on GSF innovations?
- Control innovations?
- Control failsafe?
- Position controller should be able to say we’re not achieving what we want to
- Maybe correlating poscon and ekf innov would let you tell difference between wind and estimation problems?
- Poscontrol failsafe does non-GPS-land?
- Takeoffs where vehicle was above plate in ground causing temporary but significant incorrect heading at takeoff
UTC0009 - GSoC
- Near end of submission period
- Less than 24 hours to go?
- Wording around new contributors is worrying
- Way too many Rover submissions…
UTC0015 - close