Dev Call Oct 2, 2017 2300 UTC





ChiBIOS Progress

Logging of flight mode on re-arm


Board Specific Releases

Bask Aerospace
SmartHanger David Skinner
.apfw .apj vote

what about PR #6801 UAVCAN: Vendor specific data types not dsdl-compiled

Added. It helps me find them if you add the dev call tag.

many thx, sir


Attendee count (max): 24

10:04am - Randy and Copter
3.5.3 still
Still progressing with 3.5.4
Probably not this week
Competition taking Randy’s time ATM
16, 17,18, 19, 20 of October
Find a dummy, deliver something
Several issues in CUAV which have been fixed up help with finding stuff
Particularly the focus-on-region-of-interest patch
Still have a problem with getting images timestamped with the tlogs appropriately
Timezone issues
Naming issues
Would be nice to get the camera responding the to the flight controller’s commands
So you only take photos when you need to
Might want a MAVProxy module to take the images
Only using CUAV off-line, processing a directory-full of images
15 MegaByte images
Saving is an issue, due to pausing
Multiple threads to take images might work
Developer unconference in China after that

10:13am - tridge and HeliQuad
Curtis Youngblood vehicle
Unable to fault stability in real vehicle, despite simulation issues
Collective pitch close to 0 means instability in sim
Could be a frame-rate issue tue to RealFlight lack of lock-step scheduling
High vibes
Knocked-about frame
First flight wasn’t-so-good
H-vs-X frame issue
H in real-life, X in sim!
More isolation for next flights
Needs final review and bring it in
Code is merged!
Inverted flight works
Rolls really rapidly
Loses less than three metres when flipping
But needs turning for this!
This isn’t al elegant flip
Just corrects being upside-down all-of-a-sudden
Maybe we should have another parameter which kicks in past a certain angle error so you can have a higher roll rate?
Could have something like flip mode
Ch7 opt at the moment
RCn_OPTION coming
Peter wants RCMAP to die
Replaced with RCn_OPT
MdB: ditching RCMAP is a GCS rewrite
General support for removing it
Three different variants
Probably not likely for HK to be interested in putting a real autopilot in it
Blog required to show how to do custom conversion
JP: Depends what fc is in it: the chibios work might make it relatively cheap.

10:27am - DavidSkinner
Bask Aerospace
2008 founded
Australian 3DR distributor
Red board blue board days!
Automated drone solution
Like airbotix
Out at mining sites
No-operator is where they want to be at
E.g. bird issue for 3 months of the year?
Have a box which operates every day without any intervention to scare birds bought my first APM from David :slight_smile:
Tridge: what do you need from ArduPilot?
Bask is developing box
Weather analysis
Solar charging
Need to land well
Accurate landings every single time
Precision take-offs and landings
Time over-runs?
MdB: definitely time-overruns without reboots
32-bit-microseconds wraps in 80 minutes
We survive this, obviously
32-bit milliseconds wraps at 49 days
Very rarely tested….
64-bit microseconds wraps after billions of days
Needs to start vehicles and run them for months
Any time-overruns for long-running vehicles
There are other drone-boxes around
Want these out on farms in Janary
Israelis are a bit smoke-and-mirrors
JP: These guys were working on something similar in Melbourne
They’re happy to help out with testing stuff
On-field testing really soon now
Buzz: there are developer resources in Australia!

10:42am - chibios progress update
No Sid (5am in India)
First-flight on a PH2.1 Cube
At Sid’s place in India
Gives a great comparison
Not running off the IO board ATM
Scheduling results are amazing
Really fast build!
Plan for IO
Create a library
reimplement the existing IO protocol
Would also work under NuttX
Could run the IO firmware unmodified initially
Then a new IO firmware based on chibios
Odd RPi based board Raspilot could also use it
Chibios already supports basic CAN framing
So CANBus should (in theory) work…
Transition plan… to get… from here…. To there….
Camera-triggering code
Rpm monitoring code
If we do hit significant issues with chibios on some boards we do have the option of keeping current nuttx-based setup
Doesn’t affect linux-based boards
Other targets for chibios?
Currently only stm32 on the horizon
Revo mini
Current bare-metal port is interesting but scheduling is an issue
No-OS vs Chibios
Why is the scheduling so much better?
Good question!
In-tree sensors helped a huge amount
Probably not worth digging into if we’re going to Chibios
20% vs 38% is awesome
Standard deviation is the key thing
Context switches seem to be much faster
Bringing up the peripherals one-at-a-time might show us where the issues are!
Most things are already up
Including USB and SD
FATFS library for chibios in use
Having a HAL abstraction eventually for file IO would be nice

11:00am - Logging the flight mode on rearm
Can be very confusing when analysing logs
Need patches for Copter 3.5.3 and Plane 3.8

11:05am - Green cube thread
Current need to track own branch with default params on top
Want it easy for OEMs to set default parameters
Modify .apj (or apfw) to include parameters
GCS would overwrite section in firmware before writing to board
Signature at either end
And size of reserved area
Tooling support required for all of this, of course
Not a difficult thing to add support for
8-byte signature
4-byte length
Insertion location in JSON?
Build process becomes more complex
Signature in loader file
Philip is happy to hold-off until this solution is in place
If this is soonish
Could tack it onto the end of the firmware
Would then work with existing releases
Wouldn’t need to change loader format
Making this work with old groundstations?
Unpack the bin file and tack it onto end
Mission Planner can become an apj editor in this case
So you can create a custom .apj file
OEMs will love this!
You’ve just flight thested the new defaults….

11:30am - issues
Separate max ascent/descent speeds
Little modification used
Should reuse old one
Landing already slows down and has split speeds
Land speed / land speed high
Do we default it? Parameterise it?
PR required…
Throttle behaviour bit?
Vehicles will continue after they RTL
Probably desirable behaviour anyway!
Fixed EKF-always-shows-as-OK on LCD displays
This one can go in so long as low-battery failsafe won’t be affected
Uavcan question
We need to use these messages before we spend the time doing the dsdl work
Documentation to add new messages would be awesome
What if other flight controllers want to implement the same messages?
Eugene and Pavel should weigh in…
Could be reasons to have them outside upstream uavcan
E.g. remote bus control protocol
e.g. serial bridge
Separate git repository for these makes for continuing support after a vendor has moved on
Subdirectory in fork of uavcan?
Know the patches won’t be upstreamed
But contains vendor sdl

11:53am - apfw vs apj
Previous votes was a tie
So a top-two vote
Luis will do it right now

11:58am - Randy and Partners
4 or 5 renewals came in from partners from 10
Special call out to LightWare for going (well!) beyond the required amount of partner donation!

many thx for having discussed the UAVCAN-DSDL suggestion

however, I don’t really understand what the result is, and what the take-home message for me should be

could maybe someone pl spend 1 min to make a cross in this questionare here (pl check the option which fits the situation closest):

[ ] we are not yet sure how to handle that, and there is thus no time line for when this may be resolved

[ ] we have some ideas, but it is not priority, and there is thus no time line for when this may be resolved

[ ] following UAVCAN Gui Tool’s path it is not as simple as implied, and there is thus no time line for when this may be resolved

[ ] we have reservations, and it likely will not come as suggested

[ ] that’s a fantastic extension, and it will be come tomorrow


Hi Olli, I’m going to reply over on your UAVCAN for hobbyists topic: