09:03am - AP_Notify and Position OK (non-agenda item)
Want to be able to notify the user using LEDs or LCD that position is OK
Need Notify bits to be common for all vehicles
we should move code which updates the status bits
A_notify is oriented towards direct-attached devices
Sometimes the to-user bit is external
E.g. on transmitter
No way in mavlink to tell if you have good enough position to fly in GPS mode
Unless you’re in position mode
GPS fusion flag?
Colouring on LED needs looking at
Maybe use extra bits in mavlink sensor health status flags for absolute position OK
09:10am: Copter and Randy
There will be an rc9
Leonard fixed autotune when leaned over
When it was pitched forward doing a roll pitch it would look very odd as the frame was incorrect
Earth frame rather than body-frame roll?
Fix is in master
EKF fix from Paul is also in master
Also a GPS injection fix
If you have >13 satellites
255 bytes in message is bad
May push out rc9 without some stuff and do a point release
Erle Brain crashing
Fix found Copter-3.5.0-rc7 unexpected crash on ErleBrain
DSMX on PixRacer image
Randy has bits but the transmitter isn’t working
New reports
Bouncing due to ins filter change
Reports on some ESC calibration not working ESC/Oneshot125 calibration problem in rc8
3.5 will b EKF2 by default
RealFlight testing
Rob’s weird Heli-rubber-band issue was not reproduced
But another issue was found
Rob will provide exact parameter file to Randy so Randy can attempt to reproduce
Randy added a RealFlight quad and a RealFlight tradheli into the params directory
For 3.6
Leonard is reworking Loiter
Make regular loiter look a little like poshold
If you release sticks at high speed in loiter it leans back hard, leans forward again and glides in
Looking for a smooth stop
Vehicle should stop leaning at zero speed
Test flights soon!
Khancyr is putting a bunch of Rover fixes in
With visual odometry a release would be nice!
09:27am - tridge and likely-flying / likely-indoor
EKF likes to know if the vehicle is flying
E.g. clear of magnetic anomalies
Current mechanism is not-so-good
Height of a certain number of metres
But a lot of vehicles often ALWAYS stay low
So full mag fusion wouldn’t start
Tridge has added a likely-flying flag
Takes burden off the EKF in trying to determine that
Still a guess, of course
Detecting indoor flight
EKF likes to know “probably indoors”
No “indoors detector”
Don’t want the user to have to specify
GPS velocity data to update the horizontal accelerometer offsets is a bad idea if you’re indoors
New for EKF3
Good for hot sensors
But if you have a really poor fix problems willhappen
“Indoors” really means “really bad GPS fix”
Nice f the vehicle code could say, “likely to be indoors”
Proposal:
If you have never been in a GPS-hold-mode when armed you are likely indoors
EKF’s GPS quality is already monitored, but it turns out not to be good enough
EK2 can do accelerometer scale factors whereas EK2 doesn’t
Which means if you’re not doing the horizontal offset calibration then you’re worse off under EK2
Without movement you can’t do the xy offset calibration
Other options for determining whether you’re outdoors
Might just be able to look harder at GPS quality
Height delta?
OF makes all of this goes away
09:48am - Solo update from Matt
Nothing new from Matt
09:49am - GCC warnings with MdB
GCC7 warns you about fall-through
-Wfall-through2
No-break is not an acceptable flag
Would be nice to move to -Werror
Big PR to move the the accepted version
Will add fall-through for SITL
// fall-through?
10:00am - reverse-thrust landing on weekend
Worked really well
Optera
10:00am - vectored tailsitter also flew
Until it stripped its servo
10:00am - Award
Jani accepted an award for ArduPilot
Between DJI vice and AUVSI president
2000 people from 40 different countries
Nice to get recognition within China
Tightly packed 2 days
UAV show
Really big multicopters
3.5 metre motor-to-motor
Everyone says they’ve been developing their own autopilot system
Most use PixHawk for sure
China is really pushing Shenzhen as the world capital of drones
Chinese officials say drones should be a prime part of Shenzhen
10x-fold increase coming
Real-time data is definitely going to Chinese government
5x “drone clouds” where officials and operators can see real-time where the drones are flying
Including rego number, pilot name etc etc
Altitude Angel on steroids
Down drone remotely from this interface?!
50% of MP users are from China
May need to add facilities for Chinese market?
People were really happy to have an official ArduPilot team member there to accept the award
Only 5 others received this award
Jani now has a few really good contacts in Shenzhen
DJI vice says we’re doing a great job…
DJI as a partner?
APM vs PixHawk?
No idea of numbers
Maybe MP numbers will tell us http://firmware.ap.ardupilot.org/analog/2months/
Might be something wrong with those because of the distribution
Slow firmware.ardupilot.org
Jani will look at it
Sometimes a factor of a thousand
10:20am - ArduPilot unconference in Shenzhen!
October
Wants lots of devs there
10:21am - Funding committee
A proposal is in
Not much feedback
Vote needed from Philip, tridge and Bill
Under approval limit
Approved!
Delays are a bit of an issue
Funding for Matt?
Replace dead solos?
We have a chunk of money, so need more proposals
Tridge has been working on smoothing out the reimbursement process
Nvpi interface for PayPal
Many different APIs, all are a mess
90% of SPI PayPal transactions are ArduPilot…
CSV download is borked
And was tortuous anyway
Format would change?!
New API seems borked
Old API is marked deprecated
Apparently holding a large amount of money in PayPal accounts is now a problem
10:40am - Clojure tools to convert dataflash logs into maps
Equivalent to mavlogdump but does it much, much faster
Should this be a separate repo in ArduPilot
probably
10:41am - fnoop’s graphing stuff
Awesome!
How do we make use of this?
Someone needs to ask fnoop