DEV call 06 Mar 2017 3pm PST = 06 Mar 2017 2300 UTC


  • AC3.5 update
  • Vision projects update


  • ardusoar status




  • update on the Intel Aero RTF support
  • support for different init scripts on PX4
  • unified console debugging
  • split HAL across libraries


  • ICM-20602 support
  • lidarlitev3
  • Oneshot critical fix
  • Aero Breakout


Are these calls something that people that aren’t developers can tune into to listen?

yes, everyone is welcome!
just make sure you mute yourself when not speaking

Awesome. So… I give up, is there a link to a chat / hangout / something? Or phone number? Am I missing something?
You will need to install mumble

Also please user your real name in Mumble

Installed and in! Thanks guys!

Attendee count (max): 32

10:02am - Copter and Randy
Copter 3.5rc2 should have been out last week, but it didn’t happen
AP was crashing if you flipped the vehicle (PWM out of range)
Oneshot code
Auav x2 board (looks like PixHawk mini except an IMU is off-by-90)
Will need to set board type to 15
May be able to go off USB id?
Or bootloader version?
Vendor ID and strings on USB - still using 3DR
Need to discuss what strings vendors should use and what AP will detect
Philip R has own IDs
Would be nice to inherit ID and string from the bootloader
Q about overwriting bootloader
“Dangerous” operation
Can brick boards
Perhaps use OTP?
Original 3DR implementation included IDs
Can’t be erased at all
OLED display wasn’t working on PH2
Only works on GPS port
Should probe all ports
As should all drivers
ToshibaLED does
GPS glitch reset
Not resetting after a GPS glitch
Copter would notice a 50m jump but would ignore it
EKF issue
Should be fixed
But Sub, Rover and Plane shoud check
PixRacer problems with ESP8266
Possibly narrowed down to a specific commit
Peter and tridge will run this down this-arvo
Table gives you what streams and channels can use DMA
Crossing the streams is dangerous as the low-level code blocks if you have a channel conflict
Randy would like to get GPS-blending patch in
Two types of switching logic
Hard-switch like what we have now
But now notifies control loops and the lilke
Weighted aggregation
Usually 50/50
Based on speed accuracy
Tries to ignore glitches
Will also blend for GPS altitude
blended bit to dataflash log GPB (GPS Blended)
Should have a GPS consistency check
If your GPSs disagree on location maybe disallow arming
Need to be able to inject lots of data into the GPSs to aid getting lock faster
If you have dual GPSs maybe you can copy from one to the other
RTK data sharing
So anyone within 20km can use a basestation

10:45am - Vision and Randy
Visual odometry
Visual based estimate of deltas in your position and attitude
Vision channel in gitter has a lot of stuff in it
Dan P
Orb slam
Slam light
Lighter-weight slam solution
Scripts for APSync!
Vision-based precision landing
JC’s patch to regularise timings is needed here
Mavlink to MQTT converter

10:51am - Plane and tridge
Samuel on call!
Andre not
Tridge has made a couple of little comments on the PR
Samuel has verified EKF against matlab model
Means tridge’s playing with the matrix code looks OK
Tride wants the code to restore old heading when it leaves soaring code so you can make progress across terrain
Likely to iterate on soaring over time
Hopefully in master over next couple of days
Simulation is a bit of an issue
Possibly expand SITL to have thermals
Samuel is modifying JSBSim ATM
Skyhookrc is sending tridge a 2.6m glider to test soaring on

11:03am: Lucas and RTF drone
Aero has similar microcontroller to PixHawk
Tested in stabilize, loiter and auto
And drift!
PR coming!
Adding new PixHawk-like boards could be a lot easier in waf
Maybe have --with-x and --without-y
E.g. turning px4io off
We’re getting close to flash limits again, so this will help
Lucas will split off the bits just for that today
Currently only waf builds work
Could just error-out if you try to make for Aero
Could resurrect tridge’s make wrapper
Unified console debugging is a problem
Mavlink statustext, console, straight to USB (vdprintf, bypassing ringbuffers)
Would be good to have a single debugging function across the board
Have to be configurable
E.g. BRD_ parameter space
Enumeration or bitmask as to where to send stuff
Early debug would still have to go out on console (before we have parameters)
FPGA code is still closed
But attempts are being made to open-source it

11:18am: splitting HAL into multiple bits
HAL should use libraries in libraries directory no worries
E.g. PX4 can use AP_Radio library
OF could be moved out

11:21am: Lucas/Geoff and Aero breakout
Lots of technical details required for mating with Aero board

11:23am: tridge and Invensense 20602 / lidarlite v3
‘608 is EOL
20602 is an improved sensor
New WHOAMI was sufficient
Lidarlite v3
No 13m bug reproduced yet?!
Detection between the three versions is a little screwy
So tridge added a RANGEFINDER_TYPE option for v3
Lightware interations are getting much, much better
Somewhat better than Garmin’s LidarLite

11:28am: issues time!
Un-needed servo initialisation
We do init the k_nones for a reason

Loiter radius increases based on altitude
Tom thinks we shouldn’t do this until you get really, really high
Tridge says a step change isn’t a good idea
Idea is to adjust tuning based on altitude
E.g. the L1 period changes with altitude!
True-airspeed vs apparent airspeed
MdB thinks unless we hit the max angle then we should change the radius
Interferes with auto landing
E.g. helical landing really wide
MdB gets 120m radius when 100m is requested
Option to set base pressure altitude you’re tuned for so you can scale correctly
Wouldn’t take ratio below 1
Could have option to simply disable the scaling
If goal is to keep G-forces under control, could we parameterise that?
Different axes have different limits on Plane
Guessing at this number could be problematic
And testing even worse
MdB will look at it

11:48am: Plane stages
How do we convey plane stage to GCS
Could change display based on which stage we’re in
Could use HEARTBEAT.custom_mode
Or HEARTBEAT.base_mode
Use mavlink2 and an extra field?
Mavlink’s named_value_int message

11:54am: cruise mode throttle/pitch bug?
Tridge will have to look at the logs off-line

11:57am: changing GPS lock setting default
Latency vs glitch rejection
Solo uses 1G lock
We’ve visited this before
Used to change it based on flight mode
Different settings can cause problems at high altitude
Will probably only change one vehicle type

12:08am: TC and tandem copters
TC has some tandem copters they’d like to test
Two Goblin 500s
Two Goblin 710s
Bolting those together
Want to deliver vaccines using these vehicles in PNG
6 servos (2 swashes / 120 degrees)
Tridge has a chinook in cleanflight
Will be able to get that working
Also a local person in Canberra with one
And fhedburgh
February 2015 he put a PR in
Tridge is short on time
Happy to help other people get this working
Or people will have to wait
TC will poke tridge on gitter to motivate tridge to work on it