DEV call 14 Nov 2016 3pm PST = 14 Nov 2016 2300 UTC


  • Ac3.4.2-rc2
  • and update on object avoidance

Plane / Issues



  • compasses… how to select which one you want when there are more than 3 in the system
1 Like


  • roundup of 3.4.2rc2 release
  • not many issues reported
  • some issues regarding “drifty loiter”
  • no conclusive logs
  • conclusive-looking logs should go to Paul
  • thought it might be EKF issues, but extra logging seems to preclude it
  • some discussion of changing EKF to average sensors to handle noise better
  • talk of looking at time delay from sensors to size buffers to save on memory and reduce the effects of noise (particular in relation to GPS)
  • Randy is hoping to catch on support in the next couple of days, then push out 3.4.2
  • Terrain health issues when in non-GPS mode (Randy hasn’t investigated yet)

Object Avoidance (AP_Proximity) update

  • there is now a simulator based on the Polygon fence
  • turn fence off completely but Proximity sensor will still report it
  • fixes and improvements may be PR’d today
  • ignore-zones (to ignore e.g. your vehicle’s legs)
  • discussion of possible auto-calibration of ignore zones
  • calibrate after takeoff?
  • explicit calibration?
  • possibly put in another section of FRAM
  • store into 3*uin32_t with 5 degree resolution
  • changes to the way avoidance is done; a fence point is created between sectors; seems to be a vast improvement to what is currently in master
  • talk of supporting cheaper sensors
  • Randy is keen to add support for more hardware
  • Sonar is awful (Philip)
  • mavlink sensor would be nice
  • Randy wants to add back-away to retreat from an object
  • maximise distance from everything
  • possibly like “super-simple” mode, so you turned on globally


  • segfault on Bebop
  • probably hal is not initialised before constructor runs
  • build failure on Bebop
  • “Plane: On Critical Battery, Land”
  • Michael duBreuil is working on this with the committer
  • deleting branches in current repo
  • discussion of branches vs tags
  • there are many workflows which work
  • having a release branch for two previous releases seems popular
  • release procedures will need to be updated
  • workflows will remain up to the maintainer (e.g. Francisco)
  • Tom is going to delete the easy branches
  • turbulence for SITL
  • Paul objects to adding “white-noise”; not a good model for turbulence
  • Michael says he’s wasted Paul’s time before with a bad wind model
  • Tridge points out we could use JSBSim’s
  • they are extensive
  • Paul will comment on PR
  • the videos showing SITL integration with Unreal engine are really interesting
  • tridge will be following that up
  • “Accel calibration: add better MAVLink reporting”
  • better GCS interaction during calibration (tolerate lost messages)
  • need to fast-track the MAVLink merge
  • ArduPilot wiki issues are piling up
  • request for people to peruse the list and try to knock one or two off
  • Git success
  • move to different repo?
  • tridge says he thinks the chatter really isn’t an issue
  • Rob thinks the inflated number of committers could be used against us for political reasons
  • resolution was leave things pretty much as is

Sensor Rework (tridge)

  • enabled by default in master
  • probably broke VR-Brain, Luke-Mike should be happy enough to fix
  • fast sampling has been turned off (filtering issues)
  • will be working with Leonard and/or Paul
  • in-sensor filtering problems
  • Much less CPU used
  • another step coming which will save ~30% of CPU
  • requires all SPI drivers to run in FIFO mode
  • lsm303d work-around must be disabled; another workaround required
  • more clipping is likely be evident to the user as they are not being averaged away by the sensor-smarts
  • tridge will be working with Paul to work out some of the weird effects with using fast sampling
    requests more testing - check to see if board boots(!)
  • external sensors still work
  • e.g. airspeed sensors
  • v3 lidar funding request might be coming
  • found and fixed ms5611 baro bug

Adaptive Control

  • Ryan’s been doing hte maths, tridge the testing
  • wiped an aircraft out on the weekend
  • looking good now, needs testing on more aircraft to refine parameters

Philip hardware update

  • PixHawk2 is shipping!
  • no reports of anything going wrong
  • question as to how to pick the correct compasses as there are now a couple of external compasses
  • need to choose a replacement compass as current part 5983 is going EOL

Thanks for report ! It is very valuable for people who can’t attend to the meeting !

1 Like