DEV call 03 Apr 2017 4pm PST = 03 Apr 2017 2300 UTC


Top Contributor of the Month Nominations


  • AC3.5 update
  • vision update
  • update on Solo-on-master project
  • copter fast loop latency,


*Tailsitter progress (first flights successful!)
*airspeed improvements


  • Rover 3.2.0Beta1 is coming out today




  • i2c digitial filtering,
  • RPM on any pin,
  • UAVCAN progress,
  • MindPX - supporting these boards

Attendee count (max): 28 (and late-staying-crowd)

9:02am - Maker Faire with Philip
19th of May
Intel want to pad out their big tent
They would like to demo Edison + Cube2 working together at Maker Faire
Rover cruising around the booth using intel RealSense + joule
With Copter following above it
UWB indoor positioning system
IR Lock precision on Copter to follow Rover
Need some people to attend with Philip!
Funding available!
Up to 4 people
Philip, Gene (Hex), Dan Pollock
Few extra people on booth not an issue if they are local
UWB beacons don’t currently create a fence
They could!
Randy offers to organise that
Schwag would also be good
Can be sold
Chatter will be on main mailing list
Could use visual following instead of IRLock to show the companion computer off a bit
Maybe go with a boring STM32-based simple drone flying

9:15am - Contributor of the Month award
Need to remember the APMplanner2 graphing patches!
Currently have 7 from tags and APMplanner2 graphing
Soaring code will be nominated✓&q=label%3A"Top%20Contribution"%20closed%3A2017-03-01..2017-04-01
Tridge will be putting together a google form
With appropriate application of expletives
9 in total

9:25am: Copter with Randy
Rc4 is just about due
A few issues to fix

Still no progress on the performance issue
He increased logging for us and tried to reproduce on the bench but failed
Fast-loop-latency bug fixed
we weren’t using latest gyro samples
Should now be lowest-latency of gyros-to-outputs we’ve ever had
One-shot pulses are now all nicely aligned

9:35am: tridge and AirSpeed improvements
MdB’s hard work and persistence on this!
Significant improvement for some users who’ve had bad airspeed
Bugs in both digital and analogue domain
I2c devices on stm32
I2c sucks (Philip)
Needs testing on Aero
Apologies to Lucas for putting it in without that testing, but it is an excellent fix…
Oversampling at 15
All 15 samples have to be at new value

9:43am: tridge and tailsitter support
Mark W gets prize for first tailsitter flight!
First transition was taken by someone else
Mark had bad weather
Rudder mix was initially bad
User found and fixed bug himself
No video yet
No vectored-thrust tailsitters yet
“Belly sitter”?
Paul R has a few ideas on flight dynamics
Let body of vehicle “flap” in wind

High-wind trying to descend - too much lift
Throttle almost to zero to come down
Then you lack control
Probably need to lose your position holding to bring it down
Also transitioning from a low height might work
Back-transitions at shoulder height is what Mark is after

9:45am: tridge and other frames
A bunch more flights on tiltrotors
Tricopter tiltrotor quadplane
Really nice

9:49am: Rover with Craig
Release due out today
Once Grant gets back from car breakdown issues

09:50: Jacob and Sub
do_repeat_relay wasn’t working
Tridge both caused the bug and merged the fix
Qgroundcontrol are working on code to do motor test
Sliders for testing
Motor numbering issues

Add functionality so you can reference by servo output number
Alternate proposal to use upper bits
Or another parameter to describe the schema used for the index
Use parameter 5
0, 1 or 2
0 for frame order (autopilot default)
1 motor order
2 frame order
There’s currently a big merge from ArduPilot/mavlink to mavlink/mavlink
ArduSub did first mission at depth
Accoustic precision location system
GPS on boat
Array of sonar stuff on boat read by sub
Correlating between the two
ROV on map on GCS
Maybe could re-use the beacon EKF instead?
Would need a low-latency link
Their system will be like an RTK GPS
Other systems are hundreds-of-thousands
YouTube video out tomorrow!

10:01am: Matt and solo-on-master
Small issue with the way Solo was retrieving home location
Fixed already
Every now and then the gimbal goes limp
Seemingly at random
Including just on boot
Will do this if it doesn’t get updates over MAVLink
Need logs!
Matt will get for Paul
Radians to centidegrees conversion?
Probably unrelated
Could be a timing issue
Wakes up if you use the Solex app
Probably a race condition somewhere
The LEDs also need help
Issue found
USE parameter is default to zero
Easy fix
tones need help
Maybe add an I-am-a-Solo parameter
Frame type parameter sets
For defaults
Maybe default it based on presence of OreoLED
No EKF3 on Solo ATM
It flies after you’re calibrated
But can’t calibrate
Need -rc4 out before another flight
Could be memory?!
Paul confirms this could be a real problem
Bootloader on new cubes is newer
Could cause issues
MP seems to work well for uploading firmware
Should just be able to modify the Python script used for updating the firmware
ESC issue is real!
Need the tuning!
Really need to document it really, really well
NDA issues preclude various people from talking freely about the ESC issue
The patch in Solo really screws up Solo’s capabilities
So a hardware fix is a much batter solution
E.g. Green Cube

10:25am: tridge and RPM
We’ve been limited to one RPM monitor
Flying-field fix
Autorotation support for Copter is the idea
Hardware-timer issue in NuttX was the cause for that limit
Fix demonstrates how to do interrupt handlers on pins
20us accuracy
We rarely block interrupts now

10:29am: compass-offsets patch
Hatch magnets on planes
New parameter to boost the max compass offsets

10:31am: tridge and UAVcan patch
Thanks to Francisco for comprehensive review!
2 full days to do the review
Eugene is probably out testing it right now
Patch should be merged in next few days
Very close!
Process has been wonderful for the UAVcan stuff, tridge is happy
Headers are a bit odd, but done that way to save flash space
Darned templates
PR is big enough to kill browsers
390+ comments
Jani’s Uavcan dev boards are on the way to Eugene and tridge
Rapid development from here on!
Remote uart, i2c and SPI
Chunk of work to do on that
There’s some friction with the CANbus-as-simple-transport approach
JC will be testing it with Philip and JC’s ESCs

10:37am: Luis and MindPX board
Smaller than PixRacer
Would be nice to have support
Roland would like to see support
427 without io
Tridge suspects it will be a quick port
Probably just fmuv2
Cables needed with boards!
Tridge has a PR for enabling/disabling IO
Good for really small drones
Philip also has something small and awesome and cheap coming!
Vibration isolation
Like suspension on car prevents isolation going to IMU
Natural frequency has to be away from the vibration frequencies and preferably below that of the propellers
clear/dirty chassis would be a good solution for a very small quad
Use arms as vibration dampening
CF not a good material from that point of view
Really nice documentation available
Including all schematics
Complies with HW licenses - nice!
JST-GH connectors
Dronecode-compliant connectors

10:51am: Vision update
Paul and Randy and Kai have been working together
Z-camera is generating visual odometry
Position and attitude
Being pumped into the EKF
Integrated in such a way that it can be used in conjunction with OF, GPS, rangefinder
Not mutually exclusive
Relative position estimate
Not being put in as an absolute position fix
Should be good for navigating in confined spaces with lots of obstacles
Also creating a depth map
Timing is an issue
10ms of timing accuracy would be a good start!
We could run mavlink processing a lot faster
Maybe run it in a different thread
Rather intrusive change
Current design is memory-optimized
That was really for AVR
We now have more free memory
Could have a ringbuffer of messages instead
And add a timestamp, perhaps

11:06am: sensorhead
Outline of concept
Reservations about reliability
Continuing to run AP on the stm32 is important
Phase2 for fallback mechanism
Phase1 is custom sketch
Needs to be done as a GSoC project
Share sensor-data between STM32 autopilots?
Closely-coupled RTK
Transport-independent protocol
So it can be done over canbus
Initially it will be uart