Dev Call Nov 13, 2017 2300 UTC

Development Team
1

Copter

  • 3.5.4 release progress
  • New Loiter (from Leonard)
  • Flight mode object change for Copter (from Peter Barker)

Plane

  • 3.8.3 release

General

  • apj tool

issues





2

Attendee count (max): 20

Pre-agenda:

23:05 - Randy and Copter

  • 3.5.4

  • Its out and tested

  • Compass IST8310 is not working

  • Is this behind the minimise-features macro?

  • Jordi explicitly said v2

  • 4 fewer compass drivers on v2

  • Right on the edge for being able to put v3 onto bodgy old processors

  • MO: v3 gets loaded onto board that have 2mb flash. if its less, v2

  • Anything with this on it should be able to take the v3 firmware!

  • https://store.mrobotics.io/mRo-GPS-u-Blox-Neo-M8N-HMC5983-Compass-p/mro-gps003-mr.htm

  • LIS3mdl also doesn’t seem to work

  • V2 and v4 firmwares

  • Problem that the devs don’t appear to have the modern GPS/compass units to test on

  • Leonard has been working on a new loiter controller

  • Has reworked it since last tested

  • Should have a very direct feel

  • Directly controls lean angles

  • Feels like alt hold

  • Still has a brake in it

  • Actively comes to a stop

  • Should be able to replace poshold with this

  • But Randy’s not confident everyone would be happy with that

  • So he won’t be trying to force people off poshold

  • Maintains a desired velocity/position

  • Acceleration control

  • Extra prediction for accelerations

  • Actively estimating where vehicle will be based on desired acceleration

  • Old new loiter never made it in

  • SkyRocket guys didn’t like it that much

  • Randy liked it!

  • Available in Leonard’s repo

  • https://github.com/lthall/Leonard_ardupilot/tree/20171004-new-loiter-ac353

  • Based on something quite old - August 9th

  • 2000+ commits behind…

  • Probably worked off old-new-loiter

  • Flight-mode-objects

  • [10:22 AM] (Channel) khancyr: rover is the most up-to-date now :stuck_out_tongue_winking_eye:

  • [10:22 AM] To Weekly devcall: Not for long! :stuck_out_tongue:

  • Randy’s working on top of Peter’s branch, moving some more stuff from AttitudeControl

10:22am - Rob and compass issue

  • Total field changing on long runs

  • Bug in silicon?

  • Physical effect?

  • AC3.3

  • Ancient

  • Difficult to work with logs from this old code

  • Reproduce on something more recent?

  • Difficult to do legally

  • EKF1

  • Temperature issue

  • Need to look at innovations

  • Use world magnetic model to show what sensor should give

  • Should be visible over shorter distances

10:32 - tridge and Plane 3.8.3

  • Tailsitter things

  • New transition code is working very well for a lot of people

  • Each actuator needs to have separate speed-scaling

  • Propeller-blown vs non-propellor-blown

  • Servo_n_flags?

  • https://www.youtube.com/watch?v=vblx6Uk4Ktw

  • Lots of control surfaces!

  • Inc. canards

  • Speed-scale elevons if they’ve got airflow from pops

  • Speed-scale left-elevon with lef-throttle and right-evelon with right-throttle

  • Don’t speed-scale elevons out on edge on throttle but based on airspeed

  • Maybe need another model?

  • Don’t want our parameters to be a separate programming language!

  • Hover / fast-forward-flight-hover

  • Lean plane over hard in qhover also needs speed scaling!

  • What level do we do this?

  • Saturation?

  • Overall-throttle speed scaling is already in

  • Throttle + airspeed and control surface speed scaling will be a major feature of next version

  • Next major release will be 3.9

  • Dual airspeed support

  • More point releases (3.8.4) may happen before then

  • depending on what’s found

  • Tilting quadcopters?

  • Wingless quadplane?

  • Better to do it in ArduCopter

  • No lifting surfaces

  • Shouldn’t be that hard to do it in Copter

  • AHRS view code

  • Using motors backend would be a lot of work

  • Better to do stabilisation in frame of rotors

  • And allow IMU to be tilted in flight

  • Servos don’t tilt servos, they tilt the IMU….

  • Will let current code all work

  • Rotating AHRS may not account for loss of control in some axes when you tilt the rotors

  • https://www.youtube.com/watch?v=vblx6Uk4Ktw

  • betaflight

  • Kk boards just do rate but supported it a long time ago

  • Amilcar’s company is up to shenanigans to do with odd airframes

  • When these things get real lifting surfaces tridge is interested

  • Or when you’ve got a subset tilting etc

  • If the motors go all the way forward then you lose roll control

  • Need control surfaces

10:47am - apj tool

  • PR is in

  • Frame-specific-firmwares done right?

  • Uses a fixed 8k of flash

  • Could make it smaller on some boards

  • 2k on v2?

  • Playing linker tricks to minimise code size proved to be difficult

  • Tools/scripts

  • Need to start using it on the servers

  • Should be good for OEMs

  • Can give customers a “binary” which sets customer params

  • E.g. Rob’s helis

  • E.g. BlueRobotics subs

  • Stable releases required before this will be useful

  • Defaults.parm requires recompiling the firmware

  • This doesn’t need it

  • Will deprecate or remove old mechanism

  • --show will dump content of 8k

  • These override the standard ardupilot defaults

  • Won’t override variables which the user has set

  • Could have a parameter which forces the variable from the 8k section?

10:55am - Rob

  • Integration of Iris Automation

  • Computer vision system

  • Collision avoidance for drones

  • Machine learning

  • Can see aeroplanes

  • Keen to work with us

  • https://www.irisonboard.com

  • Fnoop might be a good person to work with them?

  • https://www.irisonboard.com/technology/

  • Already lists APM…

  • ROS activation, apparently

  • Pierre?

  • Monocular camera

  • Rob will get them in touch with tridge and Peter

11:02am - issues

11:21am - Contribution of the Month

  • Needs somebody on point to look after it

  • JP is suggesting some nominations should occur

  • CO and BG will be nominated for the heli work

  • JP will create a google voting form for nominations

  • and an email

  • Should we limit number of times it can be won by a person?

  • Once for all time?

  • General agreement on call

  • But need a discussion on the mailing list

  • Maybe once every 2 years?

  • FF: Ever might leave us with no one to attribute it to

  • RL: wouldn’t the voting tend to favour those that haven’t won?

  • Didn’t seem to work that way

  • MdB: But in that case maybe not giving it is fine?

  • MdB: The main concern is that by virtue of being the only person elgible (ie made a change and hasn’t gotten it) isn’t always worth the prize (IE small wiki change)

11:29am - reversed parameter at boot and Copter

  • DDVP only, of course

  • Claim that there’s a safety issue

  • CO showed a good demo of it

  • Changed reversed parameter while heli was live

  • Should it not take effect ‘til reboot

  • Tridge wasn’t convinced

  • Electric planes have had this forever

  • Set reversed and the thing goes

  • May confuse setup as ESC malfunctioning on reboot is harder to diagnose than when things take effect straight away

  • Many parameters can be changed to get the same effect of this reversal

  • E.g. pwm-min

  • Would need special-case code for this functionality in the servo library

  • Maintenance cost

  • RL thinks a reboot should be required

  • He understands that there’s many ways this can happen

  • But that we can fix this one

  • Currently contained in the heli rsc stuff

  • But stuff should be moved to be in the servo-output stuff

  • BG: But the pass-through stuff requires PWM

  • Tridge: you can do pass-through

  • But better to be able to use scaled input/scaled output so the output range doesn’t have to match your input range

  • Even in manual pass-through

  • When Rob wrote it he assumed off-pwm of 1000 and 2000 as full-speed

  • No need to reversing etc etc

  • Tridge current stuff handles much more than 1000-2000 outputs

  • RPM feedback from motors is what we really need

  • RL wants to retain the option of reversing ESC

  • 1500 for disarm

  • Rebooting can be problematic

  • Will start spinning

  • [11:45 AM] (Channel) JW: The pulse gets cut short

  • [11:45 AM] (Channel) JW: and interpreted as a shorter pulse

  • [11:46 AM] To Weekly devcall: @Jaconb did your patches “fix” that?

  • [11:46 AM] (Channel) JW: As far as I can tell

  • [11:46 AM] To Weekly devcall: Well we should damned well get that merged, then… maybe I can convince Dale to test that out at the field.undefinedundefinedundefinedundefined