MSL Reference Architecture�(MRA)

Why MRA: Hardware standardization efforts and contributions

Hardware

• TURTLE 5k
• 4-cam omni-vision Falcons�

Different team, similar hardware

• Omni-wheels
• Omni-vision
• MSL kicker
• Ball-handling

​

​

​

​

​

​

​

ROP: https://www.techunited.nl/media/images/ROP_RoboCup_Symposium_2013_Open_Hardware_Track.pdf

No standard but a lot of commonality in hardware of MSL teams

Why MRA: Software standardization efforts and contributions

Software

• RTDB – Real-Time DataBase
• Most teams, different scope
• Ref box + MSL logging protocol
• Mixed-Team protocol
• Falcons, Robotsports
• Simulator
• Gazebo based simulator (Nubot)
• Common Simulator: Tech United
• ROS/ROS2
• Used by some of the teams
• RIF code on Robotsports hardware
• rsopenapi

​

A new MSL team needs:

- 3-4 years before it can play according MSL rules

- Few more years to become a top team

Other leagues: �- Soccer simulation (SSIM) �- Standard platform league (SPL)

​

Software is published after each RoboCup;�Starting point for the next RoboCup. �(build on top of, stacking innovations)

Limited re-use of software in MSL in last 20 years

MSL Reference Architecture

The 11 published software designs of MSL teams �in 2022 have commonality in the software design.

​

The classic Sense, Think, Act blocks can be divided in

multiple smaller components which can be easier

to share between teams.

​

This is the basis for the MSL Reference Architecture (MRA)

​

MSL workshop 2022 was trigger for kickoff of the inter-team MRA workgroup�

​

​

​

Goals for MSL Reference Architecture

Increasing development speed of MSL teams

​

• Jump start for new teams
• Reduce time needed for first game
• Faster competitive and applying to all the MSL rules

​

• Reuse proven software parts from each other
• Focus on own strength instead of doing everything ⇒ more innovation.
• Increase performance in general ⇒ Balanced games, challenging the top teams

​

Draft design based on 11 MSL SW architectures

Component decomposition

• 30 general components
• 8 hardware interfaces

​

​

Refbox Computer

Sideline Computer

World Model

Behavior

Actions

Teammate

Hardware

Design is draft:�actual implementation will define interfaces.

​

Nice theory, but …

how does it work?

MRA components

​

Components are the basis of MRA

​

• Components have a standardized interface.�
• Component dependencies need to be managed.�
• Components are interchangeable.

​

​

MRA aspects

• A standardized component design
• Interface
• Semantics
• Using protobuf for now�

​

• MRA repository as marketplace for MRA components�

​

• Implementation and publish as MRA component

​

RobotSport using MRA::GetBall - First impression

• Refactor a bit to split behavior in two matching MRA components
• MRA::GetBall
• MRA::VelocityControl �
• Resulted in implicit reviewing current design → bugs found�
• MRA setup is feasible
• incremental migration
• middleware independent

​

Robotsports::MoveToBall

RobotSport::Motion

MRA::Falcons::�GetBall

MRA::Falcons::

VelocityControl

velocity (RCS)

velocity (RCS)

position (FCS)

RobotSport::Motion

MSL Reference Architecture - recommendations

Engagement is needed !!!�

• Start with discussion platform to define MRA
• Monthly Discord call
• Start MRA marketplace to exchange code
• Add defining documents to discuss
• Add your own strong components for others to use
• Pick other components for your own use
• Embed MRA in MSL community
• Stimulate participation by getting scoring points

Thank you! Questions?

end – backup slides

Why MRA: MSL status 2014

source: RoboCup MSL - History, Accomplishments, Current Status and Challenges Ahead - Robin Soetens et al.

Draft MRA system context diagram

Scope all robot related software

• robot software
• sideline software

Standardized component design

MRA Concept is inspired by

FMI/FMU

FMU is a standard for system integration

- FMU represents the simulation model

- FMI communication interface between models.

Standardized interface design

Each component has several interfaces

• Input
• Output
• Parameters (vary per implementation)
• State (vary per implementation)

​

All interfaces are using ProtoBuf (created by Google)

• widely used in the world, very mature and well supported
• multi-language support (more portable than C structs)
• built-in compatibility for adding or omitting elements
• easy json conversion to support testing (test-data in json)

​

​

How to integrate an MRA component

Inside Team repository (connect to MRA component):

• Replace existing implementation with call to MRA library
• Refactor to match with MRA component interfaces
• Test in simulator and on robot

​

Inside MRA repository (for new components only):

1. create component interface ({Input,Params,Output}.proto)
2. Generate skeleton files
3. implement tick.cpp
4. bonus: implement test.cpp

MRA platform: a central place for MRA components

For now: https://github.com/janfeitsma/MRA-prototype

​

Contains components as well as standardization(s)

​

​

Introduction of MRA usage

• MRA can be introduced in phases
• Starting with a limited component set and migrate towards full MRA
• Existing teams can improve by replace parts by better variants
• Existing teams can help other teams by publishing components

​

• MRA is middleware independent: can work with RTDB and/or ROS

​

MSL Reference Architecture Roadmap

Engagement is needed !!!�

• Open up marketplace as RoboCup-MSL github project
• Add the architecture diagrams to repository
• Explorer if FMI/FMU standard should be used:�packaging, multiple language support, standard tooling
• Setup releasing via artifactory
• MRA development will be market driven:
• Request : ask what you need the most
• Publish: Support teams by sharing

MRA increases sharing between MSL teams

More sharing between teams is key for the league

• Stop decreasing number of teams
• Help new teams to get faster to be ready to play
• Challenge the top teams: push to innovate

​

Robot TOULON (TDP 2022)

Falcons (2020) – Data flow

Falcons (2020) – ContROl Flow

ERSOW (2022)

NUBot (TPD-2021) – ROS-based

Why MRA: MSL status 2014 – strategy

“How are we going to keep making technical and scientific achievements, while also attracting more teams and reducing costs?

• Increase knowledge sharing
• Publications, MSL workshops, Robotic Open Platform (ROP)
• Affordable robot platform for MSL
• 5k TURTLE, Play basic soccer out of the box
• Evaluation and annual update of MSL roadmap”

​

Standardization is missing as an explicit part of the strategy�(Ton Peijnenburg @ MSL workshop 2022)

source: RoboCup MSL - History, Accomplishments, Current Status and Challenges Ahead Robin Soetens at atl.

Scientific / Engineering Challenge Criteria

• Presentation

• Novelty

• Interest for either the present or the future of the league

• Scientific/Technical complexity

• Scientific relevance for the league

• Importance of demonstrated experimental results

• Relevance of the published results� presented as a support for this challenge

​