RoboCup Building a Competitive Robotic Soccer Team
Michail G. Lagoudakis and Team Kouretes Intelligent Systems Laboratory School of Electronic and Computer Engineering Technical University of Crete Chania, Crete, Greece RoboCup: Building a Competitive Robotic Soccer Team
Sunday 17 July 2050 Spain vs Robots Soccer City Stadium
CretaMASSS 2013 Michail G. Lagoudakis Page 4 RoboCup: Building a Competitive Robotic Soccer Team Humans vs. Robots 2010
CretaMASSS 2013 Michail G. Lagoudakis Page 5 RoboCup: Building a Competitive Robotic Soccer Team Where do we stand today?
RoboCup 2011 Humanoid KidSize Semifinal (DARwin vs. Dribblers) CretaMASSS 2013 Michail G. Lagoudakis Page 6 RoboCup: Building a Competitive Robotic Soccer Team Where do we stand today?
RoboCup 2011 Humanoid AdultSize Final (DARwin vs. CIT Brains) CretaMASSS 2013 Michail G. Lagoudakis Page 7 RoboCup: Building a Competitive Robotic Soccer Team Talk Outline
RoboCup
Standard Platform League
Team Kouretes
Kouretes Research
CretaMASSS 2013 Michail G. Lagoudakis Page 8 RoboCup
Robot Champions! RoboCup: Building a Competitive Robotic Soccer Team RoboCup
RoboCup – international robotic soccer world cup – 1994: idea conceived by Hiroaki Kitano – today: RoboCup federation [ www.robocup.org ] Vision – “By the year 2050, to develop a team of fully autonomous humanoid robots that can win against the human world soccer champions ” Event – annual robotic soccer competitions in various leagues since 1997 – participation of research teams from around the world – RoboCup 2013: more than 400 teams and 3000 researchers CretaMASSS 2013 Michail G. Lagoudakis Page 10 RoboCup: Building a Competitive Robotic Soccer Team RoboCup Soccer Leagues (2013) Simulation League – 2D: 11 vs. 11 independent software agents – 3D: 11 vs. 11 software agents (humanoids)
Small Size Robot League – 5 vs. 5 small (18 cm) wheeled robots
Middle Size Robot League – 5 vs. 5 large (50 cm) wheeled robots
Standard Platform League – 5 vs. 5 Aldebaran Nao humanoid robots – past: 5 vs. 5 Sony Aibo quadruped robots
Humanoid Robot League – 5 vs. 5 custom made humanoid robots CretaMASSS 2013 Michail G. Lagoudakis Page 11 RoboCup: Building a Competitive Robotic Soccer Team Why RoboCup?
Landmark project – a project with a very ambitious goal – requires a series of technological breakthrough Examples – it took about 50 years … – … from the first flight of the Wright brothers … – … to the landing of the man on the moon – it took about 50 years … – … from the invention of the first digital computer … – … to win the human world champion in Chess Question – will robots be able to play soccer competitively in 50 years? CretaMASSS 2013 Michail G. Lagoudakis Page 12 RoboCup: Building a Competitive Robotic Soccer Team More Questions
Why humanoids? – the world is largely engineered for humans – humanoid robots can share the same environment … – … but, they first need to acquire the same skills! Why soccer? – everybody in the world knows the problem/task! – it integrates many areas of AI, ML, MAS Chess RoboCup – fully observable – partially observable – causal, static, discrete time – stochastic, dynamic, continuous time – single agent, alternating – multi agent, simultaneous CretaMASSS 2013 Michail G. Lagoudakis Page 13 RoboCup: Building a Competitive Robotic Soccer Team The Challenges of RoboCup Which skills do I need to play soccer? – can I see clearly? [ machine vision ] – what do I see and where? [ object and event recognition ] – where am I right now? [ self localization ] – what do I do next? [ decision making and planning ] – how can I help my teammates? [ multi agent coordination ] – how can I move effectively? [ motion control ] – am I sure about what is going on? [ uncertainty ] – how can I solve all these problems together? [ integration ] – how can I keep with the pace of the game? [ real time systems ] Applicability – autonomous teams of robots in dynamic environments – search and rescue, planetary exploration, surveillance, ... CretaMASSS 2013 Michail G. Lagoudakis Page 14 RoboCup: Building a Competitive Robotic Soccer Team RoboCup Workflow
[Kitano and Asada, 1998]
CretaMASSS 2013 Michail G. Lagoudakis Page 15 The Aldebaran Nao Robot
Your humanoid companion! RoboCup: Building a Competitive Robotic Soccer Team Hello! My name is Nao!
CretaMASSS 2013 Michail G. Lagoudakis Page 17 RoboCup: Building a Competitive Robotic Soccer Team The Story of Nao
Concept – low cost robot for education, research, and entertainment – a replacement robot platform for the Aibo at RoboCup Manufacturer – Aldebaran Robotics, Paris, France – started in 2005 and today counts 310 people Models – limited edition to RoboCup SPL teams at ~3,200. € since 2008 – academics edition for universities at ~10,000. € since 2010 – full market release for the public at ~12,000. € since 2012
CretaMASSS 2013 Michail G. Lagoudakis Page 18 RoboCup: Building a Competitive Robotic Soccer Team The Evolution of Nao
AL-01/02/03 2005 AL-04 AL-05.a 2006 AL-05.b 2007 2007 V2 2008 V3 V3.2/3.3 2009 V4 2010 2012 CretaMASSS 2013 Michail G. Lagoudakis Page 19 RoboCup: Building a Competitive Robotic Soccer Team Nao Parts
CretaMASSS 2013 Michail G. Lagoudakis Page 20 RoboCup: Building a Competitive Robotic Soccer Team Nao Hardware
Computer – Intel Atom Z530 1.6GHz processor – 1GB SDRAM memory, 2 GB Flash memory – Embedded GNU Linux (Gentoo) operating system – 802.11g wireless LAN and ethernet LAN Electronics – chest board for controlling devices (ARM, 10 ms cycle) – USB, I2C, RS 485 links – 6 cell, Lithium Ion, 21.6V, 2.15Ah battery Specifications – weight: 5.2 kg, height: 58 cm CretaMASSS 2013 Michail G. Lagoudakis Page 21 RoboCup: Building a Competitive Robotic Soccer Team Nao Sensors
External – 2 color CCD cameras (1280x960, 30 fps, 73°) – 2 ultrasonic sensors (0.25 m – 2.55 m) – pressure sensors (4 FSRs on each foot) – bumper sensors (1 on each foot) – auditory sensors (2 microphones) Proprioceptive – encoders on each joint (12 bit, 0.1 ο precision) – inertial unit (3 accelerometers, 2 gyroscopes) – battery level sensor
CretaMASSS 2013 Michail G. Lagoudakis Page 22 RoboCup: Building a Competitive Robotic Soccer Team Nao Actuators
Motion – head (2 degrees of freedom) – 2 arms (4 degrees of freedom) – 2 legs (5 degrees of freedom) – pelvis (1 degree of freedom) – a total of 21 degrees of freedom Light and Sound – light (RGB LEDs in the eyes, chest, feet) – light (10x2 Blue LEDs in the ears) – sound (2 speakers, 2W)
CretaMASSS 2013 Michail G. Lagoudakis Page 23 Standard Platform League
Where software makes the difference! RoboCup: Building a Competitive Robotic Soccer Team Standard Platform League (2013)
CretaMASSS 2013 Michail G. Lagoudakis Page 25 RoboCup: Building a Competitive Robotic Soccer Team Standard Platform League
Concept – all participating teams use the same robotic platform – robots with identical hardware (computer, sensors, actuators) – focus is placed exclusively on software development – promotes the development of effective algorithms Platforms – 1997 – 2008: Sony Aibo quadruped robots – 2008 – today: Aldebaran Nao biped robots Info – four legged league (4LL) [ www.tzi.de/legged ] – standard platform league (SPL) [ www.tzi.de/spl ] CretaMASSS 2013 Michail G. Lagoudakis Page 26 RoboCup: Building a Competitive Robotic Soccer Team SPL Aibo
German Open 2007: Impossibles (blue) vs. Kouretes (red) CretaMASSS 2013 Michail G. Lagoudakis Page 27 RoboCup: Building a Competitive Robotic Soccer Team SPL Nao
RoboCup 2011: Noxious Kouretes (blue) – NTU Robot PAL (red) CretaMASSS 2013 Michail G. Lagoudakis Page 28 RoboCup: Building a Competitive Robotic Soccer Team SPL Rules
Games – two halves, teams switch sides and team colors – players start on their own half of the field Penalties – 30 seconds or permanent removal from the game – player pushing, illegal defense, leaving the field, holding the ball Referees – one head referee, two assistant referees, game controller Game Controller – software for monitoring the game (time, score, penalties) – communicates information to all robots CretaMASSS 2013 Michail G. Lagoudakis Page 29 RoboCup: Building a Competitive Robotic Soccer Team Gradually Increasing Difficulty
CretaMASSS 2013 Michail G. Lagoudakis Page 30 RoboCup: Building a Competitive Robotic Soccer Team SPL Research Challenges
Perception – visual object recognition through a directed camera Localization – motion uncertainty, limited view of landmarks, perceptual aliasing Motion – legged locomotion and complex motion skills Coordination – communication, dynamic role assignment, team strategies Autonomy – self command and on board processing
CretaMASSS 2013 Michail G. Lagoudakis Page 31 Team Kouretes
www.kouretes.gr RoboCup: Building a Competitive Robotic Soccer Team Team Kouretes
China, 2008 Austria, 2009 Singapore, 2010 Turkey, 2011 Mexico, 2012 Κουρήτες [ www.kouretes.gr ] Netherlands, 2013 – RoboCup team of the Technical University of Crete (2006) – the first (and still the only) RoboCup team in Greece Members – undergraduate students of the Technical University of Crete ... – ... pursuing their diploma theses on RoboCup related topics CretaMASSS 2013 Michail G. Lagoudakis Page 33 RoboCup: Building a Competitive Robotic Soccer Team Participation in Competitions – RoboCup 2006 [ Bremen, Germany ] – German Open 2007 [ Hannover, Germany ] – RoboCup 2007 [ Atlanta, USA ] – Rome Cup 2008 [ Rome, Italy ] – RoboCup 2008 [ Suzhou, China ] – German Open 2009 [ Hannover, Germany ] – Rome Cup 2009 [ Rome, Italy ] – RoboCup 2009 [ Graz, Austria ] – Med Open 2010 [ Rome, Italy ] – RoboCup@SETN’10 [ Athens, Greece ] – RoboCup 2010 [ Singapore ] – German Open 2011 [ Magdeburg, Germany ] – Iran Open 2011 [ Tehran, Iran ] – RoboCup 2011 [ Istanbul, Turkey ] – German Open 2012 [ Magdeburg, Germany ] – Iran Open 2012 [ Tehran, Iran ] – RoboCup 2012 [ Mexico ] – AutCup 2012 [ Tehran, Iran ] – RoboCup 2013 [ Eindhoven, Netherlands ] CretaMASSS 2013 Michail G. Lagoudakis Page 34 RoboCup: Building a Competitive Robotic Soccer Team Distinctions
– 2nd place worldwide [ MSRS Simulation Challenge – RoboCup 2007 ] – top 8 worldwide [ SPL Webots Simulation – RoboCup 2008 ] – 3rd place worldwide [ SPL Nao Robots – RoboCup 2008 ] – 1st place worldwide [ SPL MSRS Simulation – RoboCup 2008 ] – 1st place [ Robotic Olympic Games, Rome Cup 2009 ] – 6th place worldwide [ SPL Webots Simulation – RoboCup 2009 ] – 2nd place worldwide [ SPL Open Challenge – RoboCup 2011 ]
CretaMASSS 2013 Michail G. Lagoudakis Page 35 RoboCup: Building a Competitive Robotic Soccer Team Scientific Contribution
1. N. Kofinas, E. Orfanoudakis, M. Lagoudakis, Complete Analytical Inverse Kinematics for NAO , ROBOTICA, Lisbon, Portugal, 2013. Best Paper Award 2. A. Topalidou Kyniazopoulou, N. Spanoudakis, and M. Lagoudakis, A CASE Tool for Robot Behavior Development , 16 th RoboCup International Symposium, Mexico, 2012. 3. A. Paraschos, N. Spanoudakis, and M. Lagoudakis, Model-Driven Behavior Specification for Robotic Teams , AAMAS, Valencia, Spain, 2012. 4. J. Pazis and M. Lagoudakis, Reinforcement Learning in Multidimensional Continuous Action Spaces , IEEE Intl Symposium on Adaptive Dynamic Programming and Reinforcement Learning, Paris, France, 2011. 5. J. Pazis and M. Lagoudakis, Binary Action Search for Learning Continuous-Action Control Policies , 26 th International Conference on Machine Learning, Montreal, Canada, 2009. 6. G. Pierris and M. Lagoudakis, An Interactive Tool for Designing Complex Robot Motion Patterns , IEEE Intl Conference on Robotics and Automation, Kobe, Japan, 2009. 7. J. Pazis and M. Lagoudakis, Learning Continuous-Action Control Policies , IEEE Intl Symposium on Adaptive Dynamic Programming and Reinforcement Learning, Nashville, USA, 2009. 8. S. Volioti and M. Lagoudakis, Histogram-Based Visual Object Recognition for the 2007 Four-Legged RoboCup League , 5th Hellenic Conference on Artificial Intelligence, Syros, 2008. 9. G. Kontes and M. Lagoudakis, Coordinated Team Play in the RoboCup Four-Legged League , 19 th IEEE International Conference on Tools with Artificial Intelligence, Patras, 2007.
CretaMASSS 2013 Michail G. Lagoudakis Page 36 RoboCup: Building a Competitive Robotic Soccer Team Completed Diploma Theses 1. Νίκος Παυλάκης, Global Game State Estimation for the RoboCup SPL , 2013 2. Γιώργος Γεωργάκης, Field Landmark Recognition and Localization for RobotStadium Soccer, 2012 3. Μαρία Καρα ήτρου, KMonitor: Global and Local State Visualization and Monitoring for SPL, 2012 4. Γεώργιος Μεθενίτης, Player Behavior and Team Strategy for the RoboCup 3D Simulation League, 2012 5. Νικόλαος Κοφινάς, Forward and Inverse Kinematics for the NAO Humanoid Robot, 2012 6. Ίρις Κυράνου, Path Planning for Nao Robots using an Egocentric Polar Occupancy Map , 2012 7. ή ητρα Αστέρω Τζανετάτου, Interleaving of Motion Skills for Humanoid Robots , 2012 8. Αγγελική Τοπαλίδου, A CASE Tool for Robot-Team Behavior-Control Development , 2012 9. Ε ανουήλ Ορφανουδάκης, Reliable Object Recognition for the RoboCup Domain , 2011 10. Αλέξανδρος Παράσχος, Monas: A Flexible Software Architecture for Robotic Agents , 2010 11. Ευάγγελος Βαζαίος, Narukom: Distributed Communication Framework for Robot Teams , 2010 12. Ελευθέριος Χατζηλάρης, Visual-Feature-based Self-Localization for Robotic Soccer , 2009 13. Ανδρέας Πανάκος, Efficient Color Recognition under Varying Illumination Conditions, 2009 14. Walid Σουλάκης, Reinforcement Learning for Robotic Walk , 2009 15. Ευστάθιος Βαφειάς, Kouretes MSRS Nao Simulation Team , 2009 16. Γεώργιος Πιερρής, Soccer Skills for Humanoid Robots , 2009 17. Ιάσων Πάζης, Learning Continuous-Action Control Policies , 2008 18. ιονυσία Χρόνη, Motion Control of a Bipedal Robot using Reinforcement Learning , 2008 19. Σουζάννα Βολιώτη, Histogram-Based Visual Object Recognition for the 4-Legged League , 2008 20. Γεώργιος Κόντες, Coordinated Team Play in the RoboCup Four-Legged League , 2007 CretaMASSS 2013 Michail G. Lagoudakis Page 37 RoboCup: Building a Competitive Robotic Soccer Team Kouretes 2013 Members
Faculty – Assoc. Prof. Michail G. Lagoudakis – Dr. Nikolaos Spanoudakis Kinematics – Nikos Kofinas Walk Engine – Stylianos Piperakis Local State Estimation – Nikolaos Kargas Global State Estimation – Nikolaos Pavlakis Coordination – Evangelos Michelioudakis CretaMASSS 2013 Michail G. Lagoudakis Page 38 Κουρήτες Research
RoboCup@TUC RoboCup: Building a Competitive Robotic Soccer Team Robot’s View of the World
CretaMASSS 2013 Michail G. Lagoudakis Page 40 RoboCup: Building a Competitive Robotic Soccer Team Monas: Software Architecture
CretaMASSS 2013 Michail G. Lagoudakis Page 41 RoboCup: Building a Competitive Robotic Soccer Team Narukom: Communication
CretaMASSS 2013 Michail G. Lagoudakis Page 42 RoboCup: Building a Competitive Robotic Soccer Team Putting Everything Together
RoboCup 2011 SPL Qualification Video CretaMASSS 2013 Michail G. Lagoudakis Page 43 RoboCup: Building a Competitive Robotic Soccer Team Putting Everything Together
RoboCup 2012 SPL Qualification Video CretaMASSS 2013 Michail G. Lagoudakis Page 44 RoboCup: Building a Competitive Robotic Soccer Team Putting Everything Together
RoboCup 2013 SPL Qualification Video CretaMASSS 2013 Michail G. Lagoudakis Page 45 RoboCup: Building a Competitive Robotic Soccer Team Team Play
CretaMASSS 2013 Michail G. Lagoudakis Page 46 RoboCup: Building a Competitive Robotic Soccer Team Global State Estimation
CretaMASSS 2013 Michail G. Lagoudakis Page 47 RoboCup: Building a Competitive Robotic Soccer Team Coordination (Brute Force)
CretaMASSS 2013 Michail G. Lagoudakis Page 48 RoboCup: Building a Competitive Robotic Soccer Team Coordination (PSO)
CretaMASSS 2013 Michail G. Lagoudakis Page 49 Conclusion
Let’s wrap up! RoboCup: Building a Competitive Robotic Soccer Team Conclusion
RoboCup – a challenging competition – spans almost all research areas of AI, ML, MAS Features – ideal domain for acquaintance with the real world – ideal domain for integration and system level thinking – ideal domain for benchmarking research approaches – ideal domain for introducing (advanced) students to robotics – ideal domain for team work and long term development Suggestion – participate, think, create, learn, …
CretaMASSS 2013 Michail G. Lagoudakis Page 51 RoboCup: Building a Competitive Robotic Soccer Team Acknowledgements
CretaMASSS 2013 Michail G. Lagoudakis Page 52 The End!
Thank you for your attention! Questions?