10:03am: Randy - Copter update
3.4.3 has been out for a few days
EKF lane switching is the most important
3.4.4 is coming
Voltage compensation
Compass calibration fixes
Out maybe in a week https://github.com/ArduPilot/ardupilot/projects/1
When do we start on 3.5?
Trying to aim for quarterly
Maybe the end of February for 3.5
Testing to start no later than beginning of January
In-tree drivers
Very few drivers not in-tree, CANBUS being the big one
Need to decide whether to leave the old drivers as an option
Should elimination of PX4 sensor drivers trigger a major release?
Px4fmu
aux outputs on PixHawk
Safety switch on pixracer
Gpios
ORB stuff
ioctls
mixers
Will be redone with simple C++ interfaces
EKF3 as default EKF?
Tridge thinks 2 should be the default, 3 hasn’t had enough flying
Until the avoid-calibration code goes in the advantages are not fully realised
May have less of an ability to recover from a bad attitude
EKF1 didn’t recover. EKF2 did recover. EKF3 might recover.
Lots of other good things for 3.5
Possibly reduced latency on rate controllers by rearranging the stack
Vision updates
OpenMV camera
Python-programmable
Maybe use for precision landing
Very small (~5cmx3cm)
Zed-camera
Object avoidance possibilities
Depth perception
Wiki updates
Have improved our ranking dramatically
Page titles are important!
From consultant https://docs.google.com/document/d/1zMG1pCg1AK1Nng_Ost8ivv1lesyTFXTkJDebJm01LpE/edit
Should we replace the ardupilot.org page with the same technology as the wiki?
Randy thinks we went this way for ease-of-integration of external feeds.
Tridge thinks it was originally for the blog-posting
10:45 AP_Sync updates with Peter
MAVProxy chews up heaps of CPU
Dronekit-la chews up lots of CPU on a Raspi
10:48 Rover updates
No Grant
Rover release due, probably today
10:49 ArduSub release with Rusty
Works with Jacob at BlueRobotics
Jacob is full-time on ArduSub
Up-to-date with master!
About to make first stable release
Several hundred ROVs released
Bringing it into master looks really nice and easy (tridge)
SITL is working but there are some outstanding problems
Visualisation
Positioning and velocity?
USBL
$15,000 -> $30,000
Would be nice to get poshold using a camera instead
Companion computer
Flask for web backend
Feeds a python app which starts and monitors things like MAVProxy
11:08 Lucas and Linux
Points out required fix in a sensor driver
Moving to US apparently takes time
11:14 tridge and HW
Using the 1MB flash on PixHawk1
Linker tricks put EKF3 into top 1MB
EKF3 stops running when USB plugged in
Apparently even new flight controllers have the CPU bug (old chips on the market)
Sadly, the bootloader we have currently refuses to load >1MB flash on the faulty chips
New scheme:
Fmuv2 build and a fmuv3 build
Fmuv3 allows all memory to be use
Problem is that then only people who have uploaded their bootlader test e.g. EKF3
So now we’re thinking that we will disable UAVCAN on v2 and v3
Perhaps on release we could nuke the PX4Firmware sensor drivers, allowing UAVCAN to return
Randy’s work on having a single copter build for all frames will help fix the build-server time problems this will cause
Plane is the closest to running out of space
Quadplane includes a lot of copter code…
PR: Get PixRacer back into the v2 build?
NuttX surgery required
Uart driver has to be configurable for its pins
Needs to probe around to work out where its FRAM is
It is possible to combine the 427s together, but it will be work
F7 is coming….
11:42 PR hardware update
Solo Cube is coming…
PX4 Pro are starting to support PixHawk
build-your-own-Iris thing coming
11:44 Rob
Putting a PixRacer on a 250 copter
Rob’s looking at open-sourcing the lot of it
Suggestion for build video
Suggestion for pre-set PIDs / parameter file