Weekly Dec Call of March 28, 2017

đź“ž Coordination Calls PX4 Autopilot Coordination
1

Meeting Minutes:

2

PX4 Dev Call - March 28 2017

What features are in release? Where is it documented?

Mathieu (new person) getting up to speed on PX4 VTOL

Goal to release next week

Want to fix Snapdragon and Aero before release

Devs need to do a better job tagging issues.

Kanban boards should be working (where?)

VTOL

  • Changes are improvements to quality of flight

  • Some safety related issues

    • Quadshute

      • Good to get in release

  • Mostly fine tuning

  • Changes not release gating

More work on master

  • Should be Matir’s focus

Are issues marked Critical flagged for release?

  • Are people paying attention to the tags?

What prep is needed for the dev call?

  • List of critical issues

  • List of release gating issues

Snapdragon

  • Testing with Christophe

  • Cross test

  • Not gating release on optic flow

Aero

  • BSP is more stable

  • Altitude hold

    • Need to resolve issue

  • Yaw happening

    • Calibration issue?

David has setup test environment for ?

Mission takeoff - copter sits on ground for a while, then it shoots up into the sky

  • Lorenz - Likely HW throttle issue

    • Need to tune altitude controller

  • What changed?

  • Need to document that new release requires new tuning or this behavior can occur

  • Lorenz: should not see a difference if you have a standard airframe

Parameters page on pixhawk needs to be updated.

  • File new issue

FastRTPS

What is the philosophy on parameters?

  • Lorenz: The less parameters, the more automation

    • Tuning 30 parameters is error prone and only experts likely to get it right

    • Core dev team needs to do the tuning

      • Seems not-scalable

    • Having fewer parameters seems likely more scalable

  • Can we have parameter nesting?

    • Need to hide complexity

Piers: Another philosophy would be that the more parameters there are, the more flexible the configuration of the platform is. For me, the question is: can the default values for those parameters be sensibly set e.g. on a group, platform or airframe type basis? Appropriate configuration of default parameter values, coupled with a good tuning guide might provide a scalable solution.

Need to announce times on when to test current stuff vs new stuff

  • Need landing page for release info

    • Hamish working on it

  • Creating schedule for predictable releases

  • Set feature freeze data

    • Use “enhancement”, “bugfix”, etc on fixes

  • Clearly we need a document on development flow, how to tag, etc.

SITL and integrating rotors

  • Lorenz: Almost done re-integration

  • PX4 ROS not being tested in CI anymore

  • Building rotors is a pain

    • Only need 2 pkgs but need to clone whole repo

    • Use existing PR

      • Use additional flag for ROS build

  • SITL Gazebo repo work

    • Causing code drift?

    • Need core plugins pushed upstream to Gazebo

    • Lorenz to follow up with OSRF

    • Are we providing what OSRF needs for testing (vs what Ardupilot gives them)

Vili working on

  • Have 2 people working (part time) on it

  • Need to work out plan for FastRTPS and QGC

    • What is the end goal?