1 of 14

RoboCup SSL Open TC/OC Meeting

Strategic Road Map and Plan Discussion

2 of 14

Event Discussion

How did the event go overall?

​

Technical challenges feedback?

​

Discussion about the future of the league and technical direction will take place next.

​

​

​

3 of 14

RoboCup SSL Open TC/OC Meeting

Strategic RoadMap and Plan Discussion

4 of 14

Things We Hear From Teams (we think)

  1. Increase the number of real and virtual games played per year

​

  • Improve average game quality

​

  • Ensure students interests (and lab funding pipelines) are healthy
    1. Publication cadence (other stuff from the trustee level?)

​

​

​

5 of 14

Guiding Principles of the SSL

  • Building Community
    • Interesting Gameplay
    • Ease of Entry
    • Educational Relevance
    • Scientific Discourse
    • Open Source

​

  • Advancing Technical Progress
    • Effective AI Progress
    • Quantifying Complex Results (AI and hardware performance)
    • Robustness to Real World Uncertainty
    • Striving for full autonomy

​

​

​

6 of 14

Guiding Principles of the SSL

  • Building Community
    • Interesting Gameplay
    • Ease of Entry
    • Educational Relevance
    • Scientific Discourse
    • Open Source

​

  • Advancing Technical Progress
    • Effective AI Progress
    • Quantifying Complex Results (AI and hardware performance)
    • Robustness to Real World Uncertainty
    • Striving for full autonomy

​

​

​

7 of 14

Guiding Principles of the SSL

  • Building Community
    • Interesting Gameplay
    • Ease of Entry
    • Educational Relevance
    • Scientific Discourse
    • Open Source

​

  • Advancing Technical Progress
    • Effective AI Progress
    • Quantifying Complex Results (AI and hardware performance)
    • Robustness to Real World Uncertainty
    • Striving for full autonomy

​

​

​

Repeated identical questions about vision and field setup, field material, etc. We have a documentation gap.

8 of 14

Guiding Principles of the SSL

  • Building Community
    • Interesting Gameplay
    • Ease of Entry
    • Educational Relevance
    • Scientific Discourse
    • Open Source

​

  • Advancing Technical Progress
    • Effective AI Progress
    • Quantifying Complex Results (AI and hardware performance)
    • Robustness to Real World Uncertainty
    • Striving for full autonomy

​

​

​

Harder to pin point…

​

We now have an awesome way to search TDPs (thanks Emiel ☺). Can we tell which TDP section actually work well universally as written? Can we reproduce? Can we tell if a team’s success is hardware or AI based?

​

Harder to improve…

9 of 14

Other Items

  • Are there other high level objectives or problems we want to highlight?
    • Avoid any highly technical specifics for now
    • Example Opening Point: “It’s difficult to do community engagement because field setup time and cost is high”

​

10 of 14

Easy of Entry Strategic Plan

Field Construction Guides

    • Camera Selection
    • Computer Selection
    • Materials
    • Cost Estimates
    • Known Issues
      • Past RoboCups
      • Other teams setups

Shared Software Guides

    • Explain what data is sent and where for league systems
      • Video Guides
      • Comprehensive Written Guides

​

Radio Recommendations

    • Data driven from future shared radio technical challenges
    • Recommended hardware and software
    • Links to open-source drivers and software bridges

How to Start a RoboCup SSL Team

11 of 14

Improve Match Quality Strategic Plan

Fundamentals of Motion

    • Control Benchmarks
      • Can be used in TDPs
    • Technical Challenges
    • Recommendations on how to write in TDPs

Trajectory Benchmarks

    • Trajectory Benchmarks
      • Can be used in TDPs
    • Technical Challenges
    • Recommendations on how to write in TDPs

Skill Benchmarks

    • Passing Technical Challenge Data
  • Benchmarks improve TDP quality and measurability of new ideas
  • Team performance improves because they can more quickly choose better performing ideas

12 of 14

Improve AI Strategic Plan

Simulation Improvements

    • Improving AI integration and overall parallelism and speed
    • Improving the disparity between simulation and real life

Machine Learning Challenges

    • Explore machine learning to replace specific behaviors or algorithms with small well defined scope
    • Series of collaborative Technical Challenges
    • Bonus points if AI policies work on multiple teams simulators or robots

Advanced AI Explorations

    • In-Game Learning
    • Provable learning rates or error bounds, particularly for in-game learning
    • Distributed and split collaborative AI
    • Multi-team collaborative AI/agents

Enable but not require teams to use more advanced AI/ML

Ensure common AI/ML problems (infrastructure, sim to real) have league guidance and/or solutions

13 of 14

Things We Hear From Teams (we think)

  1. Increase the number of real and virtual games played per year
    1. Increasing the number teams
    2. Reduce the number of games decided by uninteresting failures/shortcomings
      1. Radio, lack of experience with rules/vision etc

​

  • Improve average game quality
    1. Improve baseline motion and planning across the league

​

  1. Ensure students interests (and lab funding pipelines) are healthy
    1. Publication cadence (other stuff from the trustee level?)

​

​

​

14 of 14

Open Discussion

  • Where do we want to be as a league?
    • Pretend we have no budget/time limitations.
    • At the beginning, avoid the discussion of technical specifics of challenges or ideas (okay as exemplar problems/solutions)

​