Dev Call Jan 22, 2018 2300 UTC



handling of do-commands that don’t complete before the nav-command completes.



Funding Committee

  • Swag

Attendee count (max): 24

UTC2302 - Randy and Copter

  • Trying to get 3.5.5rc1 out
  • Having issues with travis
  • Travis needs a rebase because of the yml
  • MdB to help out
  • Here+ GPS compass orientation fix
  • Mount control in guided mode
  • OneTimePad (Dylan) added SmartRTL as a failsafe option in master
  • Leonard’s new loiter is stuck on another issue
  • Moving position control PID objects into position control ibrary
  • Quadplane PIDs are problematic as they make use of the Copter controller
  • Sub also, but Jaxxzer isn’t fussed

UTC2302 - Randy and Rover

  • 3.2 is going well
  • non-GPS navigation doesn’t work without GPS
  • Change to look at EKF rather than GPS
  • Will create a point release relaxing the mode checks
  • Two types of object avoidance in Rover ATM
  • Need to combine
  • Old sonar-based avoidance which steers away from obstacles
  • Simple stop avoidance
  • Should add a line-following mode
  • Breadcrumb mode might be next
  • TWO follow-me PRs out there ATM
  • Breadcrumb (Rover)
  • Chase-mode (Copter)

UTC2302 - Hiding things behind developer options

  • Can we rename them to “experimental options”
  • Zigzag should go in and be marked as experimental
  • Would be nice to get zigzag and breadcrumb in, even if they’re hidden behind “experimental” (Peter)
  • MdB: (I mean behaiour in master is not locked in as far as I know until its released)
  • ZigZag needs the developers to come back to answer some questions

UTC2324 - SILS and

  • Those at GSoC conference met them (Jaime?)
  • Easing project management
  • Just got more funding
  • Happy to provide free to those that can’t pay
  • Artificial intelligence that organises your issue tracker
  • Learns about your tags and auto-tags
  • Labels things as stale
  • Finds duplicates
  • Confidence level that it gets right?
  • Below 95%
  • Looking for feedback, which is why they’re approaching us
  • Adding new algorithms
  • Lots of active development
  • Some organisations are already paying to use it
  • Cost?
  • Would probably have to be unlimited users for ArduPilot
  • They’d be OK with us using it for free…
  • We would manually have to undo changes made by the system if we decided we didn’t like it
  • Bases learning on the repo itself
  • On gitter when we want to contact them
  • Sils to write an email to so we have an email contact

UTC2338 -

  • AP_Mission
  • Do-commands work while you’re flying to a waypoint
  • Do-commands take some time to run
  • If you hit the waypoint before a do-command is complete then any others you might have done before the waypoint won’t run
  • But they may have if the waypoint was further away
  • This is inconsistent
  • The PR makes you execute all do commands
  • Should we run things in parallel?
  • Doesn’t work for e.g. yaw-to-heading
  • What happens if you have a cmd-do-delay?
  • This PR would kill things like changing your camera trigger distance
  • [10:48 AM] To Weekly devcall: How about we add a cmd-do which turns on the new behaviour?
  • [10:48 AM] To Weekly devcall: And one which turns off the new behaviour.
  • [10:48 AM] (Channel) MdB: param based makes more sense IMO
  • The downsides of not having some cmd-dos execute is bad
  • Weekly devcall: Or a cmd-do which is a modifier for the next cmd-do which says it must run

UTC2351 -

  • Current behaviour is surprising
  • Battery-failsafe RTLs in auto even if you have said “land”
  • Weekly devcall: cmd-do’s on the mind - have a cmd-do which specifies the failsafe behaviour? :slight_smile:
  • [10:55 AM] (Channel) MdB: The current behaiour is actually in odds with the docs which is worse imo
  • General agreement to remove the exception

UTC2356 -

  • Touch-and-go
  • Landing abort does much the same?
  • This automates the process
  • Allows you to tune the landing process
  • [10:57 AM] (Channel) MdB: I objected unless the param is otherwise useful for automated aborts
  • [10:57 AM] (Channel) MdB:: But that was just my opinion
  • Led to questions on naming of parameters
  • List of parameter names common to different landing types
  • Parameter naming issues are hard!
  • Wiki page on this!
  • GCS don’t like people renaming parameters
  • Need to worry about future people too!
  • [11:02 AM] To Weekly devcall: I wonder if AP_Param could have aliases… shudder
  • [11:03 AM] (Channel) MdB: peter ahhh that sounds bad :stuck_out_tongue:
  • [11:03 AM] To Weekly devcall: So if a param lookup-by-name fails we check an alias list or something.
  • [11:04 AM] (Channel) MdB: get-by-name would probably work, but doesn’t play that well with most param fetch names
  • RM: THR_MAX_PWM should have been THR_PWM_MAX

UTC0007 -

  • Fourier AHRS
  • Actually a sine-filter
  • Should this go in the estimator or in the control?
  • Or parts in both?
  • Author seems to be concerned the EKF wouldn’t cope
  • Paul R is the best to look at this
  • FF: it never starts the EKF, so it might actually be looking at DCM
  • Having the EKF cope with spinning Copters has use cases outside supporting monocopter
  • Come back to this next week after Paul R has had a look

UTC0015 -

  • Needs tridge’s input
  • Going to bare metal - is this something we want to do?
  • Nathan from BetaFlight
  • MdB: No RTOS is a pretty large deal breaker from past discussions, but tridge does want to support the hardware platform
  • Inside of the PR is use of cmsys stuff from STM which is NOT GPL software
  • MdB: Tridge, Sid might be able to provide port guidance as well
  • [11:21 AM] To Weekly devcall: @Nathan if you wanted to talk about architecture / ChibiOS changes after the dev call I’m happy to do so.
  • [11:21 AM] (Channel) FF: It looks be Apache 2.0 (which I think it’s compatible with GPL 3.0):
  • [11:22 AM] To Weekly devcall: Also

UTC0023 - Olivier and GSoC

UTC0026 - Olivier and funding

  • Making hard decisions on long-standing funding requests
  • Administration of swag shop considered too high
  • Could get zazzle to sell stuff for us
  • Costs us $18, we set the sell price
  • Is this just a distraction?
  • Luis: Not just about the money. Also about the exposure
  • Have whoever wants to run this get a cut themselves?
  • Jani offers to set something up
  • Linus is doing swag in Germany - maybe he can expand globally?
  • Setting up our own is long-term project
  • Current plan is that we present dev team with going with zazzle (team vote)
  • But if Jani shows us something awesome next week then we change that plan
  • If there are better options out there than Zazzle we go that way instead
  • SemaphoreCI work is underway
  • MdB: We have gone from 2 semaphore boxes to 6

UTC0037 -

UTC0100 - corralling people using obsolete hardware

  • Have a separate place we can direct people
  • General support for that idea