Dev Call Jul 31, 2017 2300 UTC


  • Copter-3.5 release updates

Mastering Solo

Rover updates


  • Plane 3.8

Issues & PRs


  • Wiki infrastructure update (Jani – if available)
  • Marketing Alex Perez

GSOC Update

Attendee count (max): 25

09:04 - Randy and Copter

  • 3.5.1 went out yesterday
  • just the DSM fix for Spektrum
  • lots of people using 3.5
  • based on Discussion threads!
  • going fairly well
  • just one odd occurrence
  • compass switching
  • decoder in Tools for decoding i2c numbers
  • tridge@blu:~/project/UAV/APM(master)$ ./Tools/scripts/ 73225 -C
  • addresses / ids
    bus_type:I2C(1) bus:1 address:30 devtype:1(0x1) DEVTYPE_HMC5883_OLD
  • fast sampling on 9250 has an issue meaning we can’t use the compass in the 9250 without slowing the sampling
    • we disable any compass past the first in 9250s on a system
  • time to think about 3.5.2
  • uavcan?
  • flown on weekend
  • implausibly low speed accuracies
  • possibly a different unit to that we’re expecting
  • FF: Issue is just lots of parameters changed without any conversion code from 3.5 to master
  • FF: @tridge: I think the issue is the UAVCAN message isn’t well defined
  • would be nice to have this due to all of the CAN hardware coming out
  • fix for Solo LEDS coming
  • some servo out fixes
  • GPS glitching in-flight notifications
  • battery-min-volt pre-arm checks
  • landing gear parameter to say what it does at startup

09:18am - Randy and Rover

  • good progress
  • completing the movement of code from vehicle to flight mode objects
  • should we use AC_Fence in Rover?
  • Plane’s fence is in vehicle code
  • InertialNav will have to go!
  • AP_Proximity will also be going into Rover code
  • going to have a low-speed and a high-speed controller for steering
  • depends on which is the limiting factor
  • grip vs wheel-turn-limit

09:22am - Randy

  • SLAM is coming
  • building a 3D map
  • want ability to store/load the 3D model
  • including GPS coordinates
  • or passing model off to other vehicles
  • SLAM gives you an absolute position
  • can reset based on what it recognizes
  • as opposed to visual odometry, which can’t
  • vision-based based on z-cam to start off with
  • ORB2 stand-alone
  • hopefully in next few weeks
  • pp: SLAM has to be opened to ROS , otherwise its going to be too limited we need the ROS tools
  • Passing ORB into/from ROS should be doable

09:23am - Solo updates

  • many people using 3.5 on Solo
  • many using boat-mode (taking off from the high seas)
  • nothing bad has happened!
  • Green-Cube users
  • Open-sourced SoloLink
  • CA asked whether dev-team thought it would be good of SoloLink was open-sourced
  • we said yes
  • Peter Buzz and Philip have been working with 3DR to get remaining bits of Solo open-sourced
  • Buzz has an open-sourced version building now!
    • all of the building stuff is released
    • including imx6 and stm32 stuff is released
    • bootloaders released
    • all under Apache2
    • a few things not released (yet!)
    • gimbal code
    • the phone apps
    • plastics
    • repos into ArduPilot?
    • we will turf the non-standard versions of various repos where possible
    • e.g. MAVProxy
    • e.g. pymavlink
    • will withhold these meeting notes until 3DR can do a press release getting their good PR on it
    • the social media stuff that’s already been floating around has seen huge responses from people
    • green cube and 3.5 in particular
  • 15 cloned repositories
  • help needed to get all of this cleaned up!
  • we’ve been handed this stuff on a plate which is awesome!
  • plastics?
  • not really 3DR here
  • phone app discussion is next
  • back on the agenda for next week!

09:37am - Philip and Edison

  • Intel and Joule just upped-and-died
  • sad for Philip since he was working with those teams a lot
  • really cool replacement cards coming for Cube carrier boards

09:45 - tridge and Plane

  • testing on weekend didn’t include the major PR blocking next beta
  • mixer change has gone in now
  • dual-gps with canbus and ublox
  • speed accuracy
  • transition issue in quadplane
  • quadplane transitions when we don’t have an airspeed sensor
  • we weren’t putting target airspeed in correctly
  • a patch went in ot address this
  • but not quite right yet
  • imagine you have a 70m-high mission
  • you take off to 10m using VTOL and you start to climb
  • no worries
  • climb to mission height using vertical motors instead
  • the height controller thinks you don’t need more airspeed to maintain height
  • so it doesn’t throttle-up, because TECS says you have plenty of energy right now!
  • can also partially transition and never complete
  • just forcing full throttle may not work as the drag from the vertical motors is too high to meeting transition speed
  • may need to put nose down still!
  • maybe if above target altitude allow nose-down during transition
  • normally don’t want nose down during transition
  • couple of crashes during the week
  • user error for the most part
  • commas vs dots in MissionPlanner
  • European issues…
  • 4.0 MOT_THST_EXPO is bad
  • Randy thinks we should add a check to make sure this isn’t >1
  • tridge thinks this might be better in the GCS
  • negative can make sense for some sorts of motors
  • this will probably go in, low-risk change
  • problem with 3.8 not being released but some critical fixes needed on 3.7
  • e.g. integrator reset
  • tridge wants to concentrate on 3.8
  • but is happy for Tom to try to get a 3.7.2 together
  • Tom will need a list of things to go into 3.7.2
  • Next 3.8 beta would have been released except for not being flown on weekend
  • so many changes between 3.7 and 3.8 people might be reluctant to make that big jump
  • need to get older stable and betas to be available
  • MissionPlanner does magic stuff to store away vehicle binaries
  • the documentation issues are much more significant (3.7.2 vs 3.8 instructions)
  • tridge is hoping for 3.8 in next week
  • no guarantees!
  • tridge is thinking of pulling in the PX4Pro / NuttX changes
  • won’t be based on master
  • will be a lot of it, 'though
  • Luis suggests if that it is this close then it might not be worth doing 3.7.2
  • how could people help tridge hit the target?
  • flight testing?
  • go through patches to determine which patches to include
  • only feature is the quadplane transition stuff

10:22am - MdB and mags

  • magnetometer warnings in flight
  • when your primary magnetometer goes bad we stop logging them!
  • when your primary goes unhealthy we stop using any magnetometer in the EKF!
  • that’s intentional because secondary mags usually suck
  • e.g. next to power sources
  • mag should still used in hover
  • but isn’t in quadplane
  • EKF learning compass offsets is coming
  • could start to get arming errors on ground because when you start learning compass offsets in-flight the mags might be seeing different fields due to motor interference
  • could see quite different mag fields on a quadplane depending on VTOL or not
  • compassmot calibration was flawed, small patch to fix
  • correct fix includes the per-motor mag calibration in
  • and increase it to >4
  • building well is a really good idea

10:31am -

  • AP_Landing
  • letting AFS do deep-stall landing was easy
  • ask AP_Landing if it wants to handle the landing
  • it asks the backends
  • opt-in option for the user
  • would be have a LAND_SPIRAL_TERMINATE?
  • too many options?
  • allow user to specify landing type for termination?
  • AFS_TERM_ACTION should have a different value to enable this?
  • the term action was supposed to bring the vehicle down as quickly as possible with as little horizontal movement as possible
  • the magic value is important!
  • people have been seen to set AFS_TERMINATE to 1 in flight!
  • should fs-plane take a parameter for termination type?
  • heli could have an autorotate termination function?

10:42am -

  • why does satellite count matter?
  • maybe you know you should have 17 satellites!
  • the user understands satellite count
  • every GPS driver gives you count
  • users get confused with the horizontal accuracy stuff etc
  • needs question “how good a gps do you have?”
  • calculation in Kalman filter for arming
  • should be run all the time
  • not just once to see SAcc drop
  • pass/fail continuously
  • arming check should check this
  • getting GPS locks indoors can be problematic
  • when it enters failsafe and tries to rtl to land…
  • Solo satellite numbers on front screen are a problem
  • people replacing their GPSs with bodgy satellite acceptance thresholds
  • FF: Yes, move EKF GPS checks into GPS library!
  • many parameters in EKF library scale various things
  • when going into Land you should decide whether or not to do a GPS-based landing

10:55am - reluctant-to-modify contributors

  • when it doesn’t seem to accomplish anything
  • when the contributor is unwilling to take advice on changes required to bring contribution in?
  • running from conflict!
  • comment and run!
  • repeated efforts to bring stuff in?
  • two PRs for same thing?
  • this is a social issue
  • “combative”
  • we have coding standards
  • which the contributor has refused to comply with
  • general agreement is that in this case the PR should be closed as “won’t merge”
  • suggestion that a time limit be imposed

11:05am -

  • needs tridge

11:06am -

  • needs tridge
  • serial ports are important!
  • PB: Last time I spoke with tridge about this he thought we would kill the nsh console thread based on the parameter value. You’re swinging until you read your parameters.
  • PB: could have a flag file on SD which could force the console to be active
  • hardware for CAN-to-serial?
  • no firmware for it
  • funding proposal?
  • maybe ping Craft&Theory guys?
  • PB: SD may come up too late to be useful. May need to do pin-short detection or something.

11:11am -

  • deferred 'til next week

11:15am - eLab is working on a little gimbal for the RPi2 camera

  • no promises about releasing files, 'though

11:16am - Wiki

  • all major servers should be moved by end of this week
  • wiki, www and firmware will be moved during this week
1 Like

Woohoo! I was just about to sit down and spend weeks/months working out how to replicate the functionality of the smart shots from scratch in opensource. Thanks to all who made this possible :slight_smile: