"Futebol Robótico: A Ciência para além do Futebol ", Pedro Lima, 26 de Março de 2009, Ciclo de Conferências "Das Sociedades Humanas às Sociedades Artificiais" (edição 200), Instituto de Sistemas e Robótica, Instituto Superior Técnico, Lisboa
Futebol Robótico: A Ciência para além do Futebol (Pedro Lima)
1. /40 Pedro Lima, ISR/IST FUTEBOL ROBÓTICO: A CIÊNCIA PARA ALÉM DO FUTEBOL Pedro U. Lima ISR/IST Março 2009 Das SociedadesHumanas àsSociedadesArtificiais
2.
3.
4. RoboCupSoccer Research using robot football as the leitmotiv RoboCupJunior International Education project based on robots RoboCupRescue Research on Search and Rescue robots operation in disaster mitigation scenarios Other : RoboCup@Home, @Space, NanoLeague, Standard PlatformLeague International initiative, joining some of the top IA and Robotics researchers on Multi-Agent and Multi-Robot Systems worldwide Symbolic Challenge : by 2050, build a robotic team capable of defeating the human soccer world champion www.robocup.org
6. ScientificConference RoboticCompetitions (Leagues) Robot Football Middle-SizeLeague Small-SizeLeague 4-Legged League HumanoidLeague Standard PlatformLeague SimulationLeague SearchandRescue Real RobotsLeague SimulationLeague RoboCup Junior Football Dance SearchandRescue [email_address] www.robocup.org
17. ROLES are sets of behaviors constraining the robot possible actions DEFENDER ATTACKER GOALKEEPER MIDDLEWARE & ARCHITECTURE
18. Connection to X11 interface Connection to Kicker actuator Runs individual behaviours and guidance functions Vision modes Connection to Graphical interface Intra-team Comm thread Behavior Switching Blackboard control Detects robot Sopped MIDDLEWARE & ARCHITECTURE 2001 proxy halt x11 relay vision kicker machine Desired Individual Behaviour Desired Vision Method
19. MIDDLEWARE & ARCHITECTURE ATLAS: the subsystem that supports the whole system WISDOM: a relevant requirement for intelligence to be displayed CORTEX: CoORdinator, Team organizer, EXecutor MeRMaID: Multi-Robot Middleware for Intelligent Decision-making 2008
21. NAVIGATION The Navigation Problem Guidance Joint Controller Vehicle Path Planner Localization obstacles target path or trajectory joint set points (e.g., wheel velocities) joint torques (e.g., motor inputs) sensor measurements operation point posture estimate joint state feedback
22. NAVIGATION Goal: to determine thepositionandorientationof a robotw.r.t. to a worldframe, in a fast, reliableandmoderatelyaccurateway, andinthepresenceofstrongodometryproblems, so as toresetthe robot odometryestimate Natural fieldlandmarks (such as fieldlinesandwalls) are used to determinethe robot absoluteposture, fromthe apriori knowledgeofthefieldgeometry. 2000 SELF-LOCALIZATION BASED ON OMNIVISION CATADIOPTRIC SYSTEM
29. I’m “dead”, and I can’t see the ball… The ball is 1 m away from me, 15º right The bal is behind me The ball is 1.5 m away from me, 25º right COOPERATIVE PERCEPTION
30. Boxes toofar: cannotseethem Moderate certainty about boxes position Reasonable certainty about boxes position I am pretty confident about boxes position Boxes too far: I’ll better move closer COOPERATIVE PERCEPTION
31. LOCAL VISION ONLY FUSING VISION AMONG THE TEAM ROBOTS BLUE ROBOT GETS THE BALL LOCATION FROM ITS TEAMMATES MAGENTA ROBOT GETS THE BALL LOCATION FROM GREEN ROBOT, WHO DISAGREES WITH BLUE ROBOT COOPERATIVE PERCEPTION
49. HUMAN TEAM Hugo Costelha, Vasco Pires, Miguel Arroz, Nelson Ramos, Marco Barbosa, João Santos, João Estilita, João Messias, PieroPalamara, Carlos Marques, Bruno Damas, Pedro Pinheiro, MatteoTaiana, BobVecht, João Torres, Tiago Antunes, Rodrigo Ventura, Luis Toscano, Pedro Aparício, Luis Custódio, Carlos Pinto Ferreira, Pedro Lima, …
50. /40 Pedro Lima, ISR/IST FUTEBOL ROBÓTICO: A CIÊNCIA PARA ALÉM DO FUTEBOL Pedro U. Lima ISR/IST Março 2009 Das SociedadesHumanas àsSociedadesArtificiais
Editor's Notes
Repeat until convergence: E-Step: estimate the [E]xpected value of the unknown variables, given the current parameter estimate M-Step: re-estimate the distribution parameters to [M]aximize the likelihood of the data, given the expected estimates of the unknown variables