DEV call May 8 2017 2300 UTC


  • AC3.5 release progress
  • Throttle Boost


  • TVBS test flights and takeoff improvements
  • Results of sensor timing fixes


Issues & PRs

Contribution for the month for April

Google Summer of Code

Maker Faire

Attendee count (max): 26

09:03am - Contribution of the month
Need more suggestions!

09:04 - Randy and Copter
3.5rc5 pushed out based on master
A few broken things
ZuBax GPS and compass
Eugene currently fixing as part of other PR
May make difficult to back-port
Probably something new so should be able to make a small fix
One-shot ESC calibration broken
Probably a clock rate issue
If you don’t talk @100Hz all the time they can stop listening until power-cycled
We may be talking 50Hz
Luis was able to reproduce, so easy from now
Automatic calibration caused the PixRacer not to boot
Some init call seems to be missing in the ESC cal
Solo battery failure
Video + dataflash logs
Battery seemed to just stop
Telemetry from battery stopped coming in (temp/volts etc)
User returned home
Battery timed out (lack of load)
Should stay up as it is detecting the load
Seemed to be about the same as the battery timeout when you turn it on with no load
Battery shut itself off entirely (stopped providing power) while the vehicle was still in flight
User was only 3ft up (so just a couple of broken props)
Awfully coincidental with recent smbus battery changes
Perhaps we need to talk to the battery more often?
Should add a health status to the battery library
So GCS can warn of battery malfunctions
Radio failsafe and GCS failsafe not triggering when receiver disconnected
Taranis with frsky and nopulse
Others have been unable to replicate, including Randy
Minimum failsafe was above minimum PWM, which means you shouldn’t be able to arm
Vehicle was recovered
Bench-testing was unable to reproduce
Need documentation on Piksie GPS
Parameter load is slow if you connect to the GCS soon after startup
We throw the streaming request away and then the GCS does things one-by-one
Seems to be something odd going on with MP and SITL
MAVProxy doesn’t have the issue
ML is slow when connecting
There will definitely be an -rc6
Mode-changing for Heli during run-up
PR on its way from Rob
Companion computer requesting parameters and not getting them?
Mavlink 2 issue?
SERIALn_PROTOCOL equal to 2?
Typically first time GCS connects
Happens before GCS has requested anything
Requesting all parameters makes it stop
Need a full tlog
Maybe a bug in the telemetry forwarder?
Happened on stock firmware as well
Japan Innovation Challenge
Similar to the OBC
Northern Japan
Looking at CUAV’s work
Cuav isn’t a slick system
Very capable, ‘though
More than one person in a team
Remote help allowed?
OBC doesn’t allow offsite
Could do stuff over-the-internet
Would need very solid links
5 day event
Everyone flies at same time
13 teams last year
Probably more this year
An hour to find a dummy in 2km*1km elsipse
As many drones as you want
As big a team as you want
Throttle boost
For the Trinity petrol-boosted tricopter
Scaling factor between normal electric throttle out and the petrol motor
Factor of 3 means 20% on electric out is 60% on petrol
Need to try a lower-pitch prop
Alt-hold and loiter
Pure-electric at end showed really need 2 batteries in parallel
Massive increase in endurance
Can put the petrol boost on whatever frame you like
Need award for weirdest vehicle ArduPilot flies
Fun project
Generator makes it much more complicated
Doable, ‘though
Could be scaled up
But you might just go pure-petrol instead
Net win in terms of total endurance
Blog required!

09:51am - Jaime and Maker Faire
Met with Dan today
Jaime has stuff Dan has prepared
Need help with interface ROS/PixHawk
Waypoints and RTK GPSs are working
Pierre and Randy should be able to help
JC has the Copter ready
But no spare Copter
Spare should be sent!
Need the tether!!

09:56am - tridge and Plane
Sensor rewrite / major timing bug results
Extremely good results
Lsm303d is preforming just as well as the MPU6000
PixRacers second-IMU is much better
Same with PH2
Much lower innovations
Accuracy of state estimation thus significantly better
Should be able to enable the FIFO on the lsm303d says person at stm
Even better results if we can make that happen!
Fix doesn’t change CPU usage
Correct scaling on delta-velocity integration
MPU9250 w/enabled magnetometer was badly affected
Chained clocks mean we were trying to sample magnetometer at 800Hz rather than 100Hz
Ends up at 75Hz
Because of the way the scaling factors work
Still not getting 8kHz - 7760Hz
Scaling factor is learnt based on the actual rate

10:02am - TVBS airframes
That’s Ardupilot
QHOVER and QLOITER are next (that’s alt-hold and loiter in copter-parlance)
At startup motors are pointing straight-up
Non-linear-controller for the tilt of the motors
In master now
Full auto missions with transitions before next meeting, hopefully!
He’s a great builder
4 vehicles in last few weeks
Should we allow builders to receive top-contribution?
Lots of yeses on the chat
Tridge will nominate Rick for contribution-of-the-month - next?
Control change to have motors tip forward when pitch stick pushed forward in qhover
Might be better to have a different mode
Related to Eurocopter X3 style helicopter
Compound heli
Vforward_gain starts forward motor to get wings level thus at minimum-drag
Relies on velocity-controlled mode
Needs to know achieved velocity
This is reactionary which is not great
Needs to do better
Originally for a petrol motor which is slow to respond to throttle changes
There was a PR for compound helicopters
Should revisit it now we have a more flexible output system
How do subs do this? They’re based on Copter
Forward output channels
Rather than pitch
Thrust requests?
Then something else works out how to get that thrust?
Gaining height with heli
Currently can only use collective pitch for it
Fast-forward-flight can give you lift off the disk
We need to know how to use it
Current controllers don’t handle FFF well
Need to use TECS and L1 controller
May be able to do better with the current leash arrangement
Quick fix is to adjust nav speed up and down
There other effects when playing with these
Randy is testing patches to limit lean angle based on thrust

10:26am - Lucas A and RADIO and RADIO_STATUS duplicated messages
Why do we have duplicated messages here?
not generated by ArduPilot, generated by radios
Current firmware generates both
Originally we had RADIO as part of ardupilotmega because the 3DR radios got added
PX4 Firmware were originally not interested
When the radios were the message needed to move up from mega to common
RADIOs can put out either type
RADIO no longer emitted in beta firmwares
MinimOSD not working?
Maybe they were looking for RADIO?
PX4 Firmware now build with ardupilotmega.xml but we’re still stuck with the radios which generate both
RTS/CTS not documented for the SiK firmwares
PR for Wiki coming
Beta firmware is not backwards compatible
Parameters are lost
SiK radios are becoming less ubiquitous
Other WiFi
Video is important!
Working on 8266 - worth it?
Over $1000
One serial stream / one video stream

10:45am - block-forever semaphore cleanups
What are the other users of the semaphore doing?
Block-forever in frontend of compass?
Driver grabs semaphore when it is updating data structure
No bus operations, just copying data?
Chance of dying while holding that semaphore are very bad
But you could starve the frontend of data
10ms timeout on fetch?
Attempts a straight-up-get
If it fails it polls on the semaphore
No wake-up list in this scenario
So the wait-forever can have performance benefits
Granularity of sleep is way too large
Should add fast-come-first-serve for semaphores
Should wake those waiting for semaphores
Timoeut should be in microseconds not milliseconds
NuttX can give us FIFO semaphores
API change to not use 0 for block-forever warranted?
Maybe another function call for blocking forever?

11:00am - Olivier and GSoC
We’d previously announced slot count
We’ve now announced students
Sriram and Miguel both on call - congratulations!

4 winning proposals
Retrace steps
Path shortening
Making Rover a reference vehicle for Rover work

  • improvements to internal combustion and heavy-lift helicopters
    Sensorhead port
    Chinook gasser could then RTL after dropping Rover wqith autonomous
    Coding starts May 30th
    Internal governor
    770-sized helis
    Vaccine delivery in PNG
    100km range
    Students should do short blog posts
    They make a difference
    We have a GSoC category
    Important students not work alone
    help other students with their projects
    Help from community
    GSoC channel
    These meetings
    PHD student at Colombia
    Drones and IC engines
    PX4 code base but now starting with ArduPilot
    GSoC project
    SensorHead port
    Modularise ArduPilot code so sensorIO runs on STM board and offloads the complex controller tasks to companion computer (main flight controller now)
    SITL first
    Intel compute board after that with a PixHawk
    Paper on leveraging the physical properties of drone and car engine
    Extends to other cyber-physical systems
    E.g. snapshot memory of entire system and restore it every now and then to eliminate malicious input
    Inertia means vehicle stays up
    Depends on how often you reset the controller

11:15am - Contribution of the month for April
Many non-code-related things coming
GSoC coordination
Smart battery✓&q=label%3A"Top%20Contribution"%20closed%3A2017-04-01..2017-04-30
Tridge will be putting an email out with a link to a vote

1 Like