Issues & Pull Requests
Plane
Copter & Rover
GSOC
Attendee count (max): 20
UTC1102 - Allow up-to 32 servos by IamPete1 · Pull Request #16301 · ArduPilot/ardupilot · GitHub
- Allow up to 32 servos
- Tridge went through it
- Want this in in early phases of 4.3
- Tridge fixed a few places up
- Parameter servo32_enable
- Enable parameter so we don’t end up sending many extra parameters
- This worked back in March
- Changed since then
- Testing?
- NeedsTesting
- Ground testing….
- Flight testing….
- Octaquad quadplane user needs this…
- SITL support would be good
- New target vehicle?
- ComplicatedQuadPlane?
- New target vehicle?
- More digits in board serial number
- Merged
- Allow individual baro backends to be compiled out
- Only SkyViper uses the stuff ATM
- Need to ensure MSP, CAN and ExternalAHRS
- Baros in GPS kind of suck
- Exposed in wind
- Baro errors
- Exposed in wind
- More testing required
- Python3 by default on most platforms now
- Merged!
- Improve shaping of rate controlled axes
- Started with rate-shaping control
- Parameter conversion still to be done
- Impact on quadplane?
- How does the command model get into attitude controller?
- QuadPlane will need some patches
- Copter will use different time-constant-based-shaping rather than the accel-based shaping
- Sub and Blimp would also need patches
- Quadplane won’t change behavior based on this
- How does the command model get into attitude controller?
- Acceleration limit still in AC
- Fast time constant? Accel limit will still apply
- Only if you increased accel limit and then slowed down response you’d get different handling of aircraft
- PH will follow up with QuadPlane patches
- Should the command model be passed straight into the library rather than passing results from the command model object in?
- Feature Request: Add version number to firmware file name
- Break older tools?
- E.g. MissionPlanner
- Symlink?
- Simple tool to rename apj?
- Board + vehicle type + version
- Significant amount of work?
- Manifest generation should actually do really well for this
- It’d be nice, but…
- We’d start with 4.3
- Rely on manifest to get the names
UTC1151 - https://github.com/ArduPilot/ardupilot/pull/20340
- Console output can be disabled
- Merged!
- Init sub pwm output to 1500
- For discussion not merging?
- Try to come back to this later in the call
- Still needs testing
- Hitex airspeed rcin thread
- Should work but doesn’t
- The way we do heaps on ChibiOS
- Realloc stuff on heap to 0?
- Useful for scripting restart so you don’t run out of memory
- Scripting restart can allocate larger heap
- Throw everything away and start again
- Move to a single allocate call
- Single free
- Deallocate_heap_memory which subtracts the memory heap_t from heap pointer to work out what to free
UTC0002 - https://github.com/ArduPilot/ardupilot/pull/20605
- Added compass glitch detection to fix gps_glitch showing in mag anomaly
- Should we differentiate?
- Use the GSF to work out if the compass is good?
- Just say “GPS/Heading Glitch”?
- “Compass/Heading Glitch”?
- “GPS Glitch of Compass Error”
- This is a lot of code to just change a string
- The vibration aspect is covered by existing messages….
UTC0010 - AP_HAL_SITL: init ArduSub pwm_output to 1500 by clydemcqueen · Pull Request #20403 · ArduPilot/ardupilot · GitHub redux
- Doing stuff which is done in other vehicles?
- Alternative: Send invalid PWM before the vehicle is ready?
- Use 0 as a marker value?
- Hal-safetyon sort of thing?
- A few changes in SIM_Submarine
- Check for zeroes specifically?
- MergeOnCIPass
- QRTL climb to QRTL alt first in in Q mode
- Only if going to transition to forward flight
- Will tend to over-shoot as no slow-down at top of climb
- Stopped using the waypoint controller…
- Tridge will test IRL
- Bill will test and we can merge it if Bill’s happy
- Fix eedump tool
- Merged
- Correct compilation when OSD enabled
- Merged
- idle throttle for takeoff with folding props?
- TAKEOFF_IDLE_TH?
- Less keen on RC input concept
- Say there’s a problem… what now?
- We’ll need a PR for discussion
- So flesh out on the issue and then create an implementation
- Implement Lutan fuel consu tpion
- EFI fuel consumption is really good
- Merged
- Airspeed autocal fixes
- Dual airspeedsensor autocal problems
- Will be back-ported
- Safety fixes when using QGC
- Quadplane on ground having tested RC failsafe, going to a DO_LAND_START, slider changes to guided, arms and goes back to auto
- Magic slider switches to guided and then back to auto which can be a problem as the quadplane arming checks aren’t run
- Throttle suppression code now protects but isn’t perfect due to heuristic
- MIS_RESTART?
- Disarm in Copter goes back to first waypoint on disarm
- Sprung-throttle fixes already merged
- You now can’t arm quadplanes in RTL
- Reason we got into the problem is Copter’s inability to arm in auto
- QGC worked around that my using guided to do the arming checks
- Quadplane suffered because of this!
- AUTO_OPTIONS=3 is wonderful
- Fixing QGC is a good idea
- But 2 or 3 years before it isn’t a problem
- Starts-with-takeoff has to include vtol takeoff
- Vtol takeoff in copter now a synonym for takeoff
- Quadplane now will refuse change to auto if you armed in guided and haven’t done a guided takeoff
- Fundamental issue
- Three separate things
- Fly-to-here
- External control
- Takeoff-in-guided
- Guided mode/armed…
- Is it flying to here?
- Or are we waiting for a takeoff command?
- Now much stricter
- If you arm in guided mode then we ill now wait for
- disarm
- Takeoff command
- Takeoff as waypoint
- Change mode to qloiter or whatnot
- Need to test QRTL
- Significant behavioural change
- If you arm in guided mode then we ill now wait for
- Three separate things
- Testing needed
- Obscuring the core reason with the modereason change?
- E.g. batteryfailsafe reason will get squashed
- PH will look at it
UTC0058 - Plane update
- 4.2 PR pending with several fixes already covered in this call
- Probably a week of beta and then 4.2.1
UTC0059 - Rover and Copter update
- Several resolves items
- ATC slew yaw fix
- 4.2 alt_hold still not right?
- Single IMU
- Earth frame accel was coming from 1st IMU
- EKF bias correction wasn’t being applied
- Accel from RTL was coming from the accel cal
- Now it is using the bias correction
- Logs seem to show the EKF is doing a bad job of estimating the z-bias when flying in acro with high vibes
- Andy would like accel bias to not be calculated when doing acro
- Bias in one direction in acro then another in RTL!
- EKF accel bias code broken?
- Not really
- High accels means GPS lag becomes more important
- Ahrs option bit which says “don’t use EKF accel biases”?
- Rely on accel cal when you set up the vehicle
- This is not standard flight
- First IMU is disabled
- High-accel high-clipping
- [11:10 AM] Buzz: if clipping, then set this bit automatically?
- So we’ll add a bit and then look at fixing the accel bias learning
- Not common enough to warrant patch into 4.2
- D-shot blue-jay issues
- Works in master but never worked previously?
- Users says it is a regression but not considered a blocker ATM
- MatekH743 four IMUs?
- No, the hwdef just gives options
- INS_ENABLE_MASK is the probe mask
- Different variants of the same boards need different values
- USE will log but not usem ENABLE to not probe for it and EK3 parameter to not create a core
- Probing mask particularly useful for killing IMUs which spew errors
- 4.2.0 maybe next Saturday
UTC0125 - GSoC update
- Students have been selected
- Waiting for Google to get back to us
UTC0126 - close