TC Meeting Minutes 29.07.19

[johaq]

Below a summary of the TC Meeting: (Thanks to @nickswalker for taking notes)

Introductions
What went well
    Variety of tasks was nice for teams
    Open demos (teleop)
Pain points
    Difference between what was written and how competition and task are carried out
        Ex: starting outside the arenas
    Tasks underspecified
    Breadth of tasks made refereeing, organizing hard
    GitHub issues for clarification didn’t reach everyone
    Lots of time spent training object recognition on site
    Dataset standardization
Comments on pain points
    Underspecification
        Explicit aim of this year’s rulebook
        Was a drastic change
    Dataset standardization
        Will be addressed on the website in the future (@Alex)
    Research topics in tasks
        Stage 1: not meant to have research value, focused on integration of solved problems
        Stage 2: meant to be past current methods
        Ties in with better understanding purpose of the league
Purpose of @Home
    A draft statement from early on (by league founder circa 2006): Salable domestic service robot by 2050
        Aim of roadmap effort is to determine an official goal
    Hard to come up with tasks when the balancing of objectives is not clear
        Objectives: Research value, value to audience, doable by all robots, easy to officiate…
        Having clear objective would clarify balance of concerns
    Roadmap will come later next year
        Should be interesting to watch, should help a PhD student graduate, should advance domestic service robotics
        Resolved: TC will send in area roadmaps
Focus of efforts this year
    Make successful tasks (high participation, success rate) harder 
    Shorten rulebook
        Decided: Break out some part of the document, will figure out what parts can be separated
        @Home in a nutshell, LOC procedures, Rules, Networking
    Determining how to cut tasks
        Should low participation tasks be cut?
        EC: 4 tasks for stage 1 + GPSR, 2 tasks for stage 2 + EEGPSR & Restaurant
        Mandatory GPSR, EEGPSR, Restaurant
    Should we use standard objects?
        Some objects should be standardized
            Reduces on site time
            Potentially enables more advanced manipulation
    Networking handbook
        The moment you arrive to venue, should know what to do to connect to network
Streamlining decision processes
    Proposed: TC votes on GitHub issues
        EC not enthusiastic about idea, but will let TC sort out
Introducing Docker
    Strong suggestion for 2020
        Could appear on a guidelines document
    Make clear that robot interchangeability is a goal
        In the future, may only have one or two robots on site, all teams upload code
VizBox 2020
    Works as both visualization and teleop interface
        Allows uniform judging of teleoperation intervention
        Fleet robotics work in industry is going this direction as well
        Robots are expensive (HSR cost 130K, Tiago 50K euro)
        Will come with point penalty
Pepper problems
    Flakey communication with SoftBank
    Android-only support seems to be the future of the platform (NaoQi 2.9)
        No hardware level access, no ROS
        TC is encouraged to research alternatives to present to trustees if there is interest
    Hard to design tasks that work across the leagues
    EC is okay with tasks being different across leagues (change in opinion from last year)
RCF League development funding
    Should be issued in August
    Possibility of funding for travel to a league workshop 

[semaph0r]

Thanks @nickswalker ! Shall we do the notes in markdown to further improve readability and put them in to the wiki? Or are there any confidential information? These might be omitted from the wiki.

In this way, we collect all information in one place and do not spread them over a bunch of services. This might explain desicions to other which are not directly involved in the TC. Just my two cents.