Current Development & Future Work Workshop Kassel, 20/21 November 2008 Rob Janssen
Recap “By the year 2050, develop a team of fully autonomous humanoid robots that can win against the human world soccer team” Midsize league specific: Develop a Team Of Fully Autonomous Robots That Can Play Soccer
Current status Tech United Reactive, implicit mono-play Semi-fixed default roles Basic strategy Minimized communication (demand-driven) No learning Result: Predictable & Robust platform created
Short Term Goals Multi-Play A.S.A.P. (All Sense All Play) Dynamic Role Allocation Strategic Position-Based Behaviour Fast Communication Adaptive Tactical Insight How to?
Methodology World Model running on each individual TURTLE –Partially updated by Vision –Partially updated by WiFi Strategic GamePlay –Situal Awareness –Tracking and Labeling Objects –Dynamic Passing
Dynamic Passing Passer: –Global Tracking Team Players –Global Tracking Opponent Players –Needs Reproducible Shot Receiver –Fast Visual Ball Tracking –Needs “Perfect” Ballhandling/Reception Repetitive process (Receiver->Passer etc..)
Dynamic Passing
Adaptive Tactical Insight Tactics through World Model –Communication of TURTLE-status –Communication of TURTLE-position –Communication of Opponent-position (A.S.A.P. tracking) –Communication of Ball-position Learning Algorithms –Opponent Strengths –Opponent Weaknesses Framework For Strategy
Tooling “Fast” Communication (Cambada, ~5Hz) Omnivision (SmartCam/FPGA, 200Hz) Forward Cam (SmartCam/FPGA, 200Hz) Fast & Robust I/O (Beckhoff, 1000Hz) High Performance Onboard Data Processing (Beckhoff, Dual Core 2Ghz)
Ambition Combine control theory with multi- agent behavioral design. Result: SUUPPAADUUPAAA SOCCAAA PLAY!!