Dev Call Aug 3 2020

CraigElder · August 5, 2020, 2:07pm

Attendee count (max): 22

UTC1100 - Pre-Agenda

European call went well
Maybe 8 attendees
CAN discussions
OSD parameter system discussion
- Worked out a better approach for setting up the parameters
Will continue to do the calls
- ~36 hours from now

Avoid shot noise by slewing center frequency
Typing issues
Review comments left
Need to show some logging that the slewing is effective
- Add a temporary graph to show that the slewed frequency does move smoothly
- Testing evidence

Leonard was very positive
Filter was there so newer IMUs matched the older IMUs’s hardware filtering
- When using fast sampling
This will affect existing aircraft
The filter is followed up by a lot of other lower-frequency filters
- Probably won’t make a big difference
- This came in before fast-sampling
  - 188 was the filter rate
Tridge things this was probably a bad idea
If close to the noise filter this could be a problem
Probably just added phase lag
Probably such a small effect flight tests aren’t going to be useful

During normal fixed wing takeoff we turn off crosstracking so we go straight to first waypoint
Weren’t doing this in normal quadplanes
Would do some skewing; instance of plane doubling back
So don’t cross-track after takeoff
Moving platform takeoffs for Copter will also need something like this
merged…

Want to see a graph
Sas should start at zero, measured shouldn’t be
Paul has some plans around this area
Improve support for flying without airspeed sensor
- Initialise from stall speed of aircraft
Peter to create a graph and ping tridge

Need to make sure we’re using the right hash
- We’re not
- Tridge will comment
https://github.com/ArduPilot/mavlink/pull/132 ?
- Are we no longer compliant after these message field definition clarifications
- Merged it
- Surprised to see STATUSTEXT_LONG just disappear
Ardupilot one rebased

Move to sq() rather than using powf
isHeadingLinedup isn’t standard style
Random soaring video:
- https://www.youtube.com/watch?v=VmBccDfgOZU
Can be merged after the sq() changes
Why not a flight mode?
- This is kind of like avoidance
- Lift is the trigger rather than object avoidance
Stack of modes?
- So once you’re done thermalling you pop back out of it
- Copter just initialises everything as you enter a mode…
Path-planning
- two-item stack
- Original destination and oa destination
Soaring is ground avoidance?!
Some pitfalls where not setting parameters appropriately
- Could be mitigated by having separate mode
Soaring a LUA script?
- Heuristic logic could be coded nicely in LUA?

Last time there was only one issue - remove the is-flying check
Special-cased for tradheli
- Collective rather than throttle for tradheli?
- Bill said to do it this way…
MdB’s requested changes have been done
Merged!

Add airmode rc options
Want to be sure it doesn’t impact modes where it shouldn’t make a difference
- The accessors should show that’s the case
- qhover/qloiter etc shouldn’t be affected
  - At all times when using vtol mode but airmode is enabled we shouldn’t do airmode except in specific modes
How about qassist in fbwa?
Throttle suppression on the ground
- Finger-chopping avoidance mode
Need testing report on real vehicle
Need to know throttle suppression
Zero-throttle should kill motors in fbwa etc etc
- Until throttle suppression goes away
What would an adequate test be?
Henry will do independent tests
qHover on ground when this is active will cause it to try to stabilise?
- That’s unexpected
- Doesn’t make sense to do stuff in qhover as it’s a height-control not throttle-control
- Needs to be tested
Difficult to autotest as you need disturbances to trigger stabilisation stuff
- Spin-arm to spin-min will show when airmode turns on
- Use AHRS-trim-X to roll it over a little
Fbwa + some throttle + qassist = flight
If Henry comes back and says all good we can merge this one

Returning -1 in mavlink means QGC will stop whinging about low battery
That’s the way it is defined in mavlink
Initialise battery_pct and guarantee that the passed-in parameter isn’t modified if false is returned in a comment

Message received telling them where vehicle is received
It’s a good idea
LOCAL_POSITION not used by plane much
- Plane isn’t consuming the frame on the equivalent input
Outstanding for a couple of years
Might be affected by Planck with precision landing
Single origin for the entire vehicle sounds like a good idea
- Complicated by EK10 changes?
  - Probably not, it was a bugfix
- GPS glitches?
- If we aren’t armed then DCM shouldn’t set an origin
  - Don’t be too eager!
  - GPS units can be overconfident
Deferring until we can get input from MdB

Analogue RSSI pin not working on CubeOrange in the beta
- Not sure it was ever working
Not a lot of testers for 4.0.6 beta
Important fixes pending for a long time

Some reports not total investigated yet
DJI OSD stuff not in master
Rangefinder / guided mode
HMC compass missing
- Possible HW issue
Has the new compass stuff in MP been put into a release yet?
- No, only in beta ATM
- Needs to go out before new compass stuff can work properly
  - We lose the compass_primary parameter and you must use the PRI IDs instead
  - QGC might get some interface…
No blockers so far
Maybe another week
Safety fix for Heli needs to be included
Randy will get Henry some bullet points for Wiki documentation to tell users how to reprioritise their compasses if they don’t have a fancy interface on their GCS

UTC0121 - GSoC

PH has been working on other bits and pieces
Kharshit changes need a bit more work, tridge has been flying
Rishadh has a big PR about to go in which adds ability to back away from objects in Copter and Rover
- https://github.com/ArduPilot/ardupilot/pull/14819/

UTC0125 - i2c in SITL

UTC0121 - sbuf stuff in msp PR

UTC0131 - close