Dev Call Oct 14, 2019 2300 UTC

UAV Challenge Junior Team

Issues & Pull Requests




Developer Conference

1 Like

Attendee count (max): 15

UTC1102 - Round-up of the CanberraUAV Junior team efforts

  • Peter did a lot of talking here and didn’t take notes while doing that
  • Remove Solo override of althold failsafe?

UTC1119 - Japan Innovation Challenge

  • 4 years now
  • Hokkaido
  • 3 days
  • 8 teams
  • Bigger competition in the past
    • 5 days previous years
      • And 14 teams
  • Tougher challenge this year
    • Further from takeoff position
  • First day was a washout
    • Rain and wind
    • Couple of flights but nobody achieved anything
  • Search / delivery / recovery
    • Km by 0.5 km
  • Second day one other team found the dummy
  • Don’t need to find dummy to do delivery
  • Japan and team did a successful delivery
  • Third day nobody achieved anything
    • No good delivery
  • Quite a challenge this year
  • Randy had a huge team
    • 12 people!
    • Usually 6 or 7
    • Help from telemetry company
      • Who brought along 5 people
    • Tested 5 or 6 different telemetry systems
      • Zigbees
      • Skyhoppers
      • WiFi from Pi
      • Dronemapper
        • Mesh network thing
      • 168MHz LORA
    • LORA was best
    • New message interval stuff worked very, very well
      • Combined with the LUA scripting
      • Reduced telemetry costs significantly
  • Tough but fun
  • Going back again next year
  • New winch worked very well
    • Fishing-reel company
    • Daiwo winch
    • Uart interface!
    • Not fast
    • Rate controller built in
    • Release mechanism is great
      • Weight of package keeps mouth of hook closed
      • When weight comes off the pincer releases
    • Marker on rope triggering limit switch
  • Didn’t use package place
    • Next year!
  • Lidar not available as branches too close
  • Rover would have been nice!
    • Another team did
    • Delivery is 20 minutes long so a Rover won’t work for that
    • Recovery is 6 hours
  • $30,000 for finding the dummy
  • $20,000 for delivery
  • $200,000 for recovering the dummy
  • Prize money is divided
    • Delivery was split 3 ways

UTC1133 - autotests failures

  • They need to be reliable or developer time gets wasted
  • Can’t fix azure
  • Travis is weird when it comes to reruns
    • Github rebases and travis uses the rebase
    • So you have to push up more commits or rebase to get a true rerun
  • Adding an agenda item for the weekly call to discuss all autotests that have been failing for more than a week

UTC1144 -

  • Merged!

UTC1145 -

  • AP_Scripting
  • Fix uint32_t arguments
  • Will break existing set-message-interval scripts
  • Ideally waf generates this
  • So if you edit the generated file you should be careful
  • Manual editing is bad
    • Need to use generator
  • Example Lua scripts coming
  • LUA has issues with uint32_t

UTC1147 - AP_Vehicle - adding singleton

UTC1153 -

  • Type 9 (uavcan) and no GPS?
  • Will show fix type 1
    • But it should be “no gps”
  • Assumption that if you have a driver you have a GPS
    • Because we probe then instantiate
    • Not the case here
  • Allows backend to set status-nogps
    • And changes calling conventions to support that
  • MdB will merge once he’s looked at it

UTC1157 -

  • Yaw from moving-baseline GPS + ublox
  • Mostly MdB’s work, tridge’s PR
  • Two ublox GPS RTK GPSs in moving-baseline mode to give yaw
  • Working quite nicely
  • ITOW alignment issues
  • Relposned message could arrive after pvt but the yaw is fetched by the EKF at the same time as it fetches position
  • Race condition
  • Need to wait for alignment?
    • What if we don’t get the other message?
  • 700ms timeout before extrapolation can delay relposned
  • New delay queue in EKF for yaw data?
    • Fuse on different time horizon
    • Complex
  • Do we merge this as-is or wait ‘til everything is sorted?
    • MdB: bring it in
  • Will merge

UTC0006 -

  • Add MAV_FTP
  • From MdB
  • Need to talk with Peter about this
    • Send-message-interval stuff
  • Wants review
  • Normal mavlink telemetry so you can do things in-flight
    • Important for scripts
  • Pushes in foreground thread
  • QGC and MissionPlanner tested
    • Timeout issues

UTC0009 -

  • Checks for _write_fd and _read_fd should be sufficient

UTC0013 -

  • Failsafe option bits
  • Enum handling is current concern
  • Each failsafe has handler function called when failsafe triggers
  • Decision making changes
  • Action enum?
  • Try to reuse the Failsafe_Action enumeration?
  • MdB things we should extend it
  • [11:23 AM] (Channel) MdB: Adding more action enum values is totally valid for the record

UTC0025 - Plane

  • 4.0 will be out pretty soon
  • 4.0.1 will bring in master updates after that
  • Waiting for test reports for takeoff mode
  • Next week or week after

UTC0025 - vision

  • Simulated drone racing
  • Video feed @50Hz from Realflight so we can start on vision processing
  • Recognising gates is next step
  • Going to be months at least until we’re flying through gates autonomously
  • Link to junkcode directory in video
  • D3d-shot API in Windows for fast screen capture
    • Can go up to 100Hz
    • UDP JPEGs
    • Can do real-time flight control from this
  • Drone nationals demo in Canberra this weekend
    • RealFlight does rate control
      • Lag is a real problem otherwise
    • PWM outputs from ArduPilot are rate outputs
    • Stable fast flight
    • Ardupilot flight controller out via s.bus into s.bus input of betaflight and let betaflight do rate control
  • AirSim vs RaceFlight
    • Need a real racing drone which AirSim doesn’t have by default
    • AirSim would be good for the lockstep stuff
      • RealFlight already has the racing gates and stuff which make this quick for tridge to work on
      • 70 degree pitch angles aren’t something that Morse and AirSim do
        • There are better simulators out there
    • 20m/s
    • Tridge has turned up the quality recently
  • Many steps before we can do this properly
    • Tridge would love some help!
      • Writing scripts
        • E.g. recognising gates
        • Bad script, but it works
          • Hopefully someone will improve it
  • Maybe a new motor library in place of current PR?
    • Current shape allows you to switch between rate control outputs
  • This is kind of like the boost throttle
  • Will wait for Leonard to look through this

UTC0048 - Copter

  • Still no Copter-4.0 beta
    • Maybe this week
    • Vibration failsafe
    • Paul’s been working on the EKF
      • Climbrate with high vibration
    • Leonard wants controller side changed
      • Randy agrees and will work on it
    • Randy wants this out!
    • Testers! Yay!
  • Blheli causes internal error which isn’t fun
    • Delayed main loop issue
      • Tridge will look at it
  • 3.6 support
    • 3.6.10 copter falling from sky
      • Software failure
      • Watchdog told us this
      • 3.6.11 has the i2c storm protection
      • This could be another bug completely
      • Stuff in master gives us more fine-grained stuff
        • Particularly interrupt counts
    • master mitigates storm without notification
      • Logs it but doesn’t throw internal error
      • Maybe throw a statustext the first time or decimated or whatever
    • Everyone should shift from 3.6.10 to 3.6.11
  • Confident master is good

UTC0056 - Rover

UTC0100 - another compass calibration method coming!

  • For next week

UTC0101 - done

1 Like