Dev Call June 17 2019 2300 UTC



Issues & Pull Requests




1 Like

Attendee count (max): 15

UTC2301 - GSoC update - Peter Hall

UTC2301 - GSoC update - Matt Kear

  • Autorotation flight mode for tradheli
  • Blog post summarising existing flight mode suitability:
  • No existing mode suitable
    • Coming straight down breaks undercarriages
      • Unless you have a disc loading like a 3D Copter
      • Once your disc load goes up the energy remaining in the head won’t save the vehicle
      • Building airspeed improves efficiency
        • Stops ingesting own vortex
        • Can glide in ground effect for a remarkable time
        • Have a chunk of ground-speed
          • Who will navigate?
            • Pilot initially
            • and/or circling descent
            • Use rally points into the future
        • Height is critical
          • Lidar?
            • Chris is dubious
            • Large angle of attack may give erroneous local readings
            • Mount for always-point-down?
            • Use slope-of-ground from srtl database?
            • Pilot-input?
              • Flare using collective?
            • [9:23 AM] (Channel) JW: I suspect that since some autorotations are done with nose 90-degrees down, the lidar would never be pointing at the ground until after the crash.
        • Need to know airspeed for navigation
          • Bill is working on the airspeed indicator library
  • Bespoke controllers
    • Entry
    • Steady-state-glide
    • Flare
    • Touchdown
  • Headspeed controller written from scratch
    • PID controller
    • Setting Collective blade position
  • Airspeed attitude controller based on Bill’s work
  • Two controllers are currently fighting one another
    • Working with Bill and Chris
  • Hoping to test the first two stages on real hardware RSN

UTC2327 - GSoC Airsim

  • Timing is a pain!

UTC2328 - GSoC MAVProxy

  • One PR merged since last week
  • Two pending

UTC2328 -

  • Enable scripting on all builds
  • Would have to set a parameter to enable scripting
  • Would stop people having to recompile
  • Some unit descriptions added in AHRS
  • Can go in once MdB fixes Python stuff

UTC2329 -

  • Needs testing

UTC2337 - expanding array class

  • Will use in object avoidance stuff first
  • FF may be looking at it…
  • MdB has been looking at it?
  • Why not use std::vector?
    • Memory fragmentation?
    • Allocation / deallocation a lot?
    • Vector must be contiguous
  • Need to move code out of .h
    • Template instantiation issues
    • May be able to make the type opaque in the header to avoid having the code in the header
    • Type-safety at the interface level, bytes in the code itself
      • Ringbuffer / objectarray do this sort of thing
  • [9:41 AM] (Channel) MdB: I just strongly don’t like the operator overloading/safety checks
  • [9:42 AM] To Weekly devcall: Do the overloads buy us anything? Do we want to hide the fact we’re using our own array class?
  • [] is going to be quite slow on this so using it may give false impression
  • Bounds checking may be too expensive to generally do
    • Tridge thinks not having it is fine
    • But having them under SITL should be good
    • Handling the failure cases for oob would be a lot of code
  • Fragmentation makes out-of-bounds-access very bad
  • Why is chunk size not a const?
    • Will make things much faster
    • Pass value in in constructor
  • Extra typedef?
  • Why not realloc?
    • Will change to use it
    • May not have realloc
  • ExpandingArray a good name?
  • Can use this in the future for SmartRTL

UTC2359 -

  • Fix header stuff
  • Build time reductions!

UTC0002 -

  • See Randy’s issue - make sure we don’t do mission stuff unless you are armed
  • Armed may not be a good gate
  • Should change auto so you can arm on Copter

UTC0005 -

  • Remove extra check against longitude
  • Extra check can be removed as the range of the function is constrained
  • merged

UTC0009 -

  • Allows booting on boards with no baro
  • Merged

UTC0010 -

  • Tridge would like to canonicalise on using our own functions
  • ChibiOS is wrapping the platform thing back to ours
  • Not worth chasing down
  • Just grab the patch to the uavcan servers

UTC0019 -

  • Not sure why this was marked as devcal
  • Tridge can move the last-armed-ms into the HAL or into AP_Arming

UTC0026 -

  • merged

UTC0028 -

  • More information has been supplied
  • Log download in main thread is the problem
  • Mavlink processing in main thread is going to fix this
  • Trigger group mask is only updated in one place
    • Two places to call trigger groups?
    • Timer to achieve minimum rate to ESCs?
    • Min-pulse-trigger-us
  • IO could keep counter of how often it has gone into failsafe
    • Easier once we don’t need to support the nuttx-based firmware
    • Master we can do it now
      • There is a PR to move Copter to use ChibiOS IO
      • It’s in the 3.6 branch

UTC0029 -

  • Make surface tracking an object within Copter

UTC0035 -

  • Should we repoll the mode switch or should we go into the last commanded mode?
  • merged

UTC0048 -

  • Merged!

UTC0054 -

  • Improving stability covariance matrix
  • Randy to re-test
  • un-devcalled

UTC0054 - no plane updates today

UTC0056 - Copter

  • new rc in a few weeks
  • Getting back to 3.5 is problematic as they are not present in the server?
  • Ongoing complains landing detector isn’t working as well

UTC0109 - Rover

  • Setting up inputs and outputs is problematic for users
    • Randy’s going to do a video for it
  • Probably have as many boat users as car users
  • Boat - Vehicle slows down too much at corners
    • Turns more slowly

UTC0111 - Tracker

  • Continuous server fix to go in before new beta made
1 Like