UAVCAN
The winner of top contribution for March is the AP_Soaring work for
plane by Andrey Kolobov and Samuel Tabor.
Copter
Plane
Issues
Hardware
Thanks all for the input and advice on the continued Solo on Master project! These calls are great!
Attendee count (max): 30
9:01am - intel Aero and ArduPilot Flying well
9:02am: UAVCan and Francisco
It is ready to go!
worked with Eugene over last week to get it ready
A few whitespace changes in comments going in
Over 500 comments
Crashes firefox and chrome
Probably should split into a new PR @300 comments
Huge thankyou to Eugene!
This will greatly reduce the flash size
More room for other features
Separate follow-up PR which removes the old drivers
E.g. px4 baro
E.g. Px4 compass
Maybe another 80-100k to save there
Currently ~1M to 940k
Merged live on call!
Many more drivers to come
E.g. battery
To add more protocols?
Yes, something apart from uavcan could be implemented relatively easily
Uavcan is more complicated than other protocols
9:10am: Contributor of the month
Samuel and Andre for Soaring
âSpecial thanksâ awards
General acclamation
No special hardware this month
Solarblue offers to help finding sponsorship for the monthly top-contribution prize
9:15am: Randy and Copter
Copter 3.5rc4 was released on Saturday
Various important bug fixes
No bad reports from people that have tested the release
Position hold in autotune may need some minimum threshold before activating
Kept on turning aroun, hunting the wind
Need a mavlink autotune progress message
Verbose statustext messages:
https://github.com/ArduPilot/ardupilot/compare/master...peterbarker:copter-autotune-messages?expand=1
Probably want the aux-channel fix
Solo on Master
Little incident yesterday
Flying really well
Battery cell voltages will be available soon!
Lights and the like do work once the solo parameter is turned on
Slewing thing
Need a dataflash message
Matt needs some help interpreting log data
https://github.com/ArduPilot/ardupilot/blob/master/libraries/AP_TECS/AP_TECS.cpp#L1079
Gimbal
More testing will be required
With a dry Solo
Matt is working on a Wiki page for master-on-Solo
https://github.com/Pedals2Paddles/ardupilot/tree/SoloMod-Master
Vision
http://discuss.ardupilot.org/t/ardupilot-openkai-zed-for-non-gps-navigation
zed-camera -based visual odometry flight!
Camera image processing gives estimated pose deltas
EKF consumes those
Samuel D is getting it working on fixed-wing
Tridge had missed the VisionProjects Gitter channel
It was created as a subroom
Should have been created at the top level
Does best when it has obstructions in front of it
Not SLAMming yet
Means your position drifts
ZEDâs SLAM support is apparently poor
Maybe we could use v4l in SITL to help debug this stuff?
Dan P has also been working hard on visual ROS-bsaed solutions
10:03am tridge and Plane
Very exciting week for fixed-wing development
âFixed wingâ is becoming a poor term
Lots of new frame types
https://vimeo.com/212408771
https://www.youtube.com/watch?v=_ghq8OIFz1s
This was actually done with Elevons
PR is in to do it âproperlyâ
Can work purely vectored or as a mixture of elevon/vectored
Mechanically simple compared to a quadplane
Vibes were pretty bad
PixRacer can bound about a bit
Add mass and a bit of isolation
https://www.youtube.com/watch?v=koOqenMpvck&t=85s
Lots of random mixing gain parameters
Hover vs forward flight are varied
https://www.youtube.com/watch?v=_r5TEP-QHBI&t=3s
Mark W has his tailsitters going too
Variable-pitch quadcopters?
Probably handled as helicopter variants
Probably only a few hundred lines of code
Yaw at low speeds might be an issue
Taxonomy is a real problem
10:20am: bugfix in servo outputs
Started off as just a buzzing in servos
User noticed and didnât fly
They may have exploded if he had flown
Would require a strange setup to actually get his on a multirotor
Wrong rates on primary channels was the problem
Standard octoquad quadplane setup caused the issue
10:22am: XPlane 11 released
Now works in SITL (one-line patch)
Pretty
Slow
10:23am: THR_MAX
Changing default as most people do want 100% as full throttle range
So it has moved from 75% to 100%
Moâ throttle!
3.8 ArduPlane, âEven Fasterâ
10:24am: issues
https://github.com/ArduPilot/mavlink/pull/34
Tridge will look at
https://github.com/ArduPilot/ardupilot/pull/6015
Convenient for ground stations
But could have been quite a smaller patch
mavlink2 -specific
Would have been nice if it had been mavlink1
tridge will review and put t in in next hour or so
https://github.com/ArduPilot/ardupilot/pull/6015
Added to everyone but AntennaTracker
Should probably deprecate old BATTERY2 message
Solex app will have a pane you can open to show all the battery data
Serial number too would be awesome
Follow-up patch: https://github.com/ArduPilot/ardupilot/pull/6020
More enhancements: https://github.com/ArduPilot/ardupilot/issues/6029
https://github.com/ArduPilot/ardupilot/pull/3749
Should say yes-or-no to this one
Michael has issues with this one
Seemed like a feedback loop to him where you continued to specify your current target alt
Some changes will have to be passed in
Probably EKF origin
MdB got nominated!
This one will see some attention now!
Right concept, implementation needs a little work
10:50am: Pixie driver with MdB
Niels
Pixie reports HAE rather than MSL
will accept difference for the time being, but you wonât be able to blend with other GPS
Mostly good apart from that
Couple of bugs found in existing driver as part of this work
10:55am: Nolan and relay duration delay
Currently done in tenths-of-a-second
But also dependant on loop rate
https://github.com/ArduPilot/ardupilot/issues/6009
15fps camera
RAW, not a video
Need autopilot to know when a trigger has occurred, so continuous mode not useful
For overlap percentages etc
Delay inherent with PWM
Combine pwm with servo to get a level
CAM_DURATION is in tenths
Perhaps create CAM_DURATION_MS
And migrate current value
Current default is 10 tenths
Make new default 1000
Use a timer thread rather than calling from main loop
Could use a hrt thread to do this
Get microsecond accuracy
Is it required, âthough?
1ms is good enough for Nolan
But perhaps people in the future do need microsecond resolution
Variance is the issue
Current variance could be quite significant
So 10s of microseconds variance but specified in milliseconds
Machine vision would care
We only trigger on new GPS messages slowly
Do_take_picture
Can fix this / call it at 50Hz
Trying to make the camera not-the-weakest link
HW Feedback pin is critical
Microsecond accurate
Tridge will fix
11:14am: GSOC
Final week; evaluating proposals
Going to ask for 3-5 slots
4 prospective students on call
46 proposals
A lot of them are very, very good
Dashboard says 52 proposals
Some obviously for wrong project
And some not finalised
Several of the students on the call
Weâll probably ask for 5 slots
We have 6 mentors
But some are specialists
So we have a mentor bandwidth issue
We have to specify maximum/minimum slots
No idea how we try to come up with numbers for those
Mentoring is about 100 hours/student
18th is the deadline for slot-count