DEV call May 15 2017 2300 UTC


  • AC3.5 release progress
  • Multi-video update (it works but is laggy)
  • Mastering Solo Update


  • fixed climb rate in CRUISE mode,
  • fixed tricopter quadplane tail servo in fwd flight


1 Like

Attendee count (max): 34

09:03am - Matt and Mastering Solo
Notify stuff
Dynamic vs static configuration
Jaime has a patch which makes things dynamic
Want to eliminate tone differences
Want to have it sound like ArduPilot
This is likely to go in
Having issues with scripts on Solo
Would be nice to have mavlink telemetry on Solo
“Yaw issue”
Yaw and gimbal pitch
Try using an auto mission to point the gimbal
Various other possible solutions

09:26am - Randy and Copter
Rc6 went out overnight
Esc calibration fix
Safety-switch issue
Slow parameter fix
Still some issues
PixRacer RC input not working
Smart battery should report unhealthy instead of zero
ErleBrain issues
Randy will look into this
One user reports Optical Flow is resetting
3.4.6 not happening
Working fine for Randy
Lidarlite v3
Pwm input not working
13m offset still happening!
One user report with bad power
Circle mode
In SITL Copter descends
Altitude is user-controlled in Circle mode

09:37am - Randy and Wiki

09:38am - Camera streaming Daemon
Randy has worked wth this with MO.
MP can use multiple video streams using zeroconf
Pair of C920s and it all worked nicely
Apart from the lag
Apsync stuff was greased lightning in comparison
Lag could be in the player
Vlc can itself be laggy

09:47am - tridge and Plane
Bug in cruise mode
Pitch stick ignored for Marco
Tridge was unable to reproduce
Granularity problem
Delta calculated in terms of delta-cm and delta-time
QuadPlane running at 300Hz 2m/s climb rate gave a climb rate of 0.66cm/s - which got truncated to zero
Tridge decimated the loop to 10Hz
Tilt-tricopter quadplane using tail-servo
Tail servo was fully-deflecting in forward flight
Long-standing bug
Set all motors to min when in forward flight
Including tail servo
Patch to AP_Motors
Tail-servo now goes to neutral

09:52am - Jacob and Sub
5 has been tagged as stable

09:53am - issues
Use sticks to trigger calibration?
Perhaps a GESTURE_CH which then enables gestures?
Need to have a good set of gestures
Maybe a calibration flight mode
Use startup channel
Two sorts of airspeed calibration
Zero calibration vs ratio calibration
Timed pulses from sticks (left and right roll etc etc)
Like beastx
Airspeed sensor has no way of determining how good the cal is without a GCS

10:26am - issues
RTL behaviour
Stall prevention is over-conservative
Low-airspeed, tipped down, full throttle
3 seconds hard limit
Massive loss in altitude
Stick mixing could have brought it up
Perhaps mix in something to do with how under-height you are
Nice linear interpolation function in utilities for doing smooth transitions in controllers

10:54am - issues
Rangefinder broken?
Tridge will look at log

10:56am - Lucas A and disable compass
EKF gets very annoyed if it doesn’t have a compass
Compassless is possible
Point Rover North when you start
After it stops for a while it loses its heading
Can get heading data when stationary from a very fancy GPS
Marvelmind in use
But may not provide speed accuracy information
GPA dataflash message
Pozyx used to act as a GPS
Changed to use it as a beacon system
Vehicle wants to know heading, velocity
“Is it possible for the sensor data to give you the result you’re after, or isthe sensor data ambiguous?”
If the same sensor data could mean two different things then things are not observable and EKF is stuffed
Every sensor has offset and drift
Compare still and looking east vs still and looking wet
If the vehicle moves forward or backward then suddenly heading is observable
If moving slowly then you may jump around and heading becomes difficult to determine
Observability combined with gyro drift?
Rotational rate around each axis
E.g. yaw rate
Constant drift
Start at 0-degrees per second but will then drift
EKF constantly attempts to estimate the gyro drift
It need another yaw source which doesn’t drift to estimate the gyro drift
When you stop the rover EKF no longer has a yaw source
Soo has to start trusting the gyro
When you start moving again EKF needs to get confidence in the new yaw source
The GPA stuff gives the EKF confidence in the new yaw source
Generally the EKF needs to know how big an inconsistency can be expected from data, so when to discard (or de-weight) bad data
Lucas A has a highly-magnetic environment
Emlid rtk gps - you really want 2.5.2!

11:15am - Promotions with Jani
Having a stock on each continent for shipping out
Funding proposal should be put forward
Organisation could fund production of merchandise
Seeding creation of merchandise