DEV call April 24 2017 2300 UTC


  • Update


  • AC3.5 update

Vision updates


  • vectored yaw tilt quadplanes
  • quadplane fixes with Leonard
  • new elevon and vtail functions


Attendee count (max): 27

09:01am - Olivier and GSoC
Deadline was this-morning
Slots announced, nice number.
Ranked by several team members
Lots of quality proposals
Students have been selected
No announcement until Google make the announcements
Mentors have been assigned
Google looks down on one mentor for multiple students
So lots of backup mentors
Shame we didn’t get more
We should look at doing ardupilot-funding
Can’t discuss it until after the 4th of May deadline
We should send something out now
And be ready on the 4th to send offers to those that didn’t get the GSoC slot
For transparency we should explain how we went about selecting the students that got spots
Can’t start working with students until the announcement
Tom suggests that Google makes a big issue of there being a start and an end date
GSoC is a huge thing
1,400 slots
$9,000,000 in funding
Lucky we got this man slots given we’re a new project
Probably related to the number of applications we received
Few weeks before we get underway with the students
Grading was relatively independent, but close results, which is great

09:12am - Randy and Copter
Intended to put rc5 out this week, but didn’t happen
Performance issue came up
RC off means gimbal stops working
Project page on github highlighting some issues that need to be merged in
One from tridge related to RC
Pixracer LED fix is in
Vision projects
Wiki page for setting up the ZED camera
Image going up based on APSync

User can’t take off only in guided
EKF home is not set
Conversion alt-home-ekf-origin issue
Tested in enRoute office
Walking around, and it worked beautifully
Plane only flies with GPS
So probably doesn’t need the fix for this at the moment
But sub and rover should get it
Bug in EKF-origin-changing in flight
We assumed it didn’t
Attempt in EKF2 and EKF3 to compensate for barometer drift
Using GPS hieght
There will be a switch to turn this off
Many places assume it doesn’t move
Going to have to track them down

09:28am - Patrick and OpenMV camera
Precision loiter video
Extension of balloon finder
Comparing against C920 ODroid system
14-17 frames/second
100ms latency
Gimbal is on the way
Experimenting with tags in the next few days
Lots of good help from ArduPilot devs
Lots of feedback after diydrones video, which was great

09:31am - tridge and plane
e-Flight Convergence!
Flew it with ArduPilot very successfully
We’ve supported tiltrotor quadplanes for a long time
But this one uses tilt-rototrs for yaw
Mechanically simple because one less servo
Don’t need to tilt rear rotor
There has previously been a failure caused by tilt on rear motor
RTF quadplane
~AU$500 delivered to Canberra
Not much payload
Installing autopilot was relatively easy to do
Tridge will do a page to allow people to reproduce it
Tridge wrote code at airfield…
First time tridge has flown a tilt-rotor
He’s flown tilt-wings before
More work in time to improve
Battery doesn’t last too long
Would probably take a bigger battery
25% throttle gave 16m/s
Luis says that 4S can be used in the standard config
Perhaps the ESCs were replaced, ‘though?
Speed was discussed a bit
No differential thrust for yaw in fixed wing
No flips yet!
This aircraft allows tridge to test many things on the one plane
Vectored yaw patch was tiny
servo -channel stuff made it easy
A channel is wasted because the tri-code still outputs rear tilt
Can’t fly this on pure-arducopter at the moment because the yaw-stuff was done outside AP_Motors

09:43am - tridge - Leonard+QuadPlanes
Leonard has an X8 quadplane
Shown off at Avalon airshow
Folding props
Overspinning was a problem
But fixed with a parameter
Low current in forward flight
Has already found half-a-dozen bugs
Tridge is really looking forward to lots more of the same
Leonard seems to be happy with the QuadPlane code now
Wants to combine PIDs
Really pushing it around
Found more bugs in last couple of weeks than any other single creation

09:46am - tridge / new elevon and v-tail support
New flaperon support
Using new servo output stuff
Three ways of setting up elevons in master
Tridge will probably remove the deprecated one
And possibly the second method, but wants to migrate configs to new method

09:47am - 3.8 plane release?
Perhaps after the PixRacer fix goes in
Needs documentation before it can be done
Fixed wing can now fly a huge variety of aircraft
Release video could be nice!
Another random vehicle (tail-sitter):

09:49am - interesting vehicle:
~AU$160 delivered

09:56am - sub update with Jacob
3 or 4 reports with 3.5 rc
Working on Companion stuff
Going to lose 99% of people if they have to log in via ssh
Motortest PR still pending
Tridge will help get this in today

10:02am - issues list
Fetching of serial numbers from batteries
Solo and non-Solo ones
No logical mapping from marked serial number to technical serial number
Using parameter values to store information about serial number and battery capacity
But it gets reset on fresh batteries
Battery capacity problems
When value not being updated
Will use a flag value

Randy thinks we really should keep it simple

10:14am - Parameters in general
Olivier asks if more parameters is worse if we have good defaults
Randy thinks it should “just work”
People get intimidated by too many parameters and may wander off to find a “simpler system”
Knock-on impacts downstream for GCS
E.g. more dropdowns in calibration screen
Scares people off
That screen is already scary
Randy is really just concerned with user-interface
Currently have 1004 parameters in Copter

10:16am - more isues list
Poshold bug
Already merged in
Waiting for Rob to do pull requests for other modes
Will get to it next week, hopefully
Compile-time switches for Solo
Jaime’s PR makes it run-time
Francisco has some issues
Do we really need a separate tone-file for Solo?
Probably not…
Do we proceed with Jaime’s approach?
Jaime want to make Notify devices dynamic
OreoLED makes for a large binary
But currently v3 will flash onto v2
Solo would interrogate OreoLEDs at boot and would then firmware-update them if they need it
So why are we carrying the binary blob around if we’re not using it?
We shouldn’t
Should install OreoLED for v3 unconditionally
Should be a PR in for this
Jaime has made a patch to control it via parameter

10:35am - DMA-to-a-pin
Buzz has this working
D-shot support is getting closer!

10:37am - Donuts
Extensive testing is required
ArduDonut required

10:38am - FllyteBase has made a post about being ArduPilot compatible
Was Bill using these at one point?
Now multi-system
Not just ODroid

10:40am - adding sub-stages using mavlink2
Problem with making ArduPilot-specific
May have issues with QGC if we try to use the base mode for other things