robot teams - JNRR 2013, Annecy
Transcription
robot teams - JNRR 2013, Annecy
Interaction homme-robot Tour d’horizon et défis Rachid Alami LAAS-CNRS Toulouse - France JNRR – Annecy – 16 Octobre 2013 A shift in the perspective: From a single robot to humanrobot teams 2 1 Autonomous Robotics n a dream and a challenge n The robot was alone n n n n n Very capable Very intelligent You give it a goal/task a It does the job … end it is ready for a new adventure hero The collaborative robot n New abilities n Far more challenging and subtle … 2 n n Robots will be among us Robots will have to be able to do something useful 3 n Robots will have to be safe Robots will have to plan (and adapt) their activities with us or simply in our presence n Robots will need models and algorithms n Models of us n 4 When it comes to Human-machine interaction We have a precursor 10 5 « Modern Times » n n n n n n n n Shot between 1932 and August 1935 Première : Februray 5, 1936 Proximal Interaction Initiative Taking Adaptation to human Pertinence of choices / decisions Acceptability issues Reliability – Fault tolerance An added parameter: the human n n … Besides design choices, it is necessary to endow the robot with the ability to take explicitly into account the presence of humans Re-visit the robot decisional capabilities 12 6 Questions for a robot which collaborates with humans: what, who, where, when, how? Do, Observe, Decide/plan, Communicate Integrative approach for a robot that acts in interaction with humans n Work on Collaborative / Interactive task achievement n n n n based on a study of human-robot interaction inspired from Joint activity / Teamwork concretized as a set of robot decisional abilities Work: Inspiration and Collaboration n n n n n Cohen P. R., Levesque H. J. (1991), Tambe (Teamwork), H. Clark (Joint Activities, Dialogue)� Bratman (1992). Shared cooperative activity. Tomasello M. Warneken F. et al (2005 - ..) Knoblich G. Sebanz N. et al. (2009 - ..) Pacherie, E. (2012 - ..) 7 Topics n n n n n n n n Social signal processing / Situation assessment Production de signaux sociaux (boucles sensori-motrices multi-modales, rythmes ..) HR Physical interaction Situated dialogue Theory of Mind / Belief Management Proactivity and turn taking Acceptability issues Intentional actions and motions Integrative approach for a robot that acts in interaction with humans n Work on Collaborative / Interactive task achievement n n n n based on a study of human-robot interaction inspired from Joint activity / Teamwork concretized as a set of robot decisional abilities is progressively producing a coherent basis for Joint Human-Robot Activity 8 List of co-authors Rachid Alami, Samir Alili, Gérard Bailly, Michael Beetz, Ludovic Brethes, Maxime Cottret, Raja Chatila, Aurélie Clodic, Patrick Danes, Kerstin Dautenhan, Xavier Dollat, Peter Dominey, Frédéric Elisei, Isabelle Ferrane, Sarah Fleury, Martin Haegele, Matthieu Herrb, Mamoun Gharbi, Mokhtar Gharbi, Guillaume Infantes, Felix Ingrand, Ken Koay, Madhava Krishna, Thibaut Kruse, Jens Kubacki, Jean-Paul Laumond, Christian Lemaire, Frédéric Lerasle, Séverin Lemaignan, Efrain Lopez Damian, Jerome Manhes, Philippe Marcoul, Luis Marin, Jim Mainprice, Paulo Menezes, Vincent Montreuil, Christopher Nehaniv, Christopher Parlitz, Amit Pandey, Raquel Ros Espinoza, Daniel Sidobre, Thierry Siméon, Akin Sisbot, Mick Walters, Felix Warneken, Matthieu Warnier, Britta Wrede, Sara Woods. A task-oriented architecture for an interactive robot � Task-Oriented: How to perform a task, in presence or in interaction with humans, in the best possible way � � � � H&R Sharing Task and Space Efficiency Safety Acceptability Intentionality � Plan-Based: Planning and OnLine Deliberation � � � � Reasoning Anticipation Pro-active behaviour Theory of Mind – Predicting and reasoning about human activity and mental state 9 Decisional ingredients for an Interactive Autonomous Assistants A Control Architecture for a Cognitive Assistant Human-Robot Joint Action Affordances Perspective Taking Human Aware Navigation Shared Plans: Human Aware Task Planning Human Aware Manipulation RIS Adapting motion dynamics “Soft” Motion 19 Who can do what with which object with which effort. Put Video (Manipulability_Graph_Validation_Short) Put Video (Manipulability_Graph_Validation_short_sce_2) 10 Now let us clean the table (or put the tools in the box) + Who can do what with which object with which effort. Who can do what for whom, with which effort and where. Put Video (PR2_clean_table_affordance_graph_shorter_2.) Affordance Graph Robot dealing with HR shared activity n Robot maintains a (symbolic) state: n n n n n Robot elaborates plans involving H & R n n n n tracking of objects visible to it Tracking of human posture and hands visibility and reachability of objects w/r human Reachability of objects wiith respect to iself Computes a plan involving (potentially) both itself and the human Human action is selected when necessary Robot computes also grasps, placements and motions Robot executes plans and monitors H activity n n n By performing its own actions Asking human to achieve some actions Re-plans when necessary: state different from what is expected 11 Robot dealing with HR shared activity n Robot maintains a (symbolic) state: n n n n n Robot elaborates plans involving H & R n n n n Tracking of objects visible to it Tracking of human posture and hands Visibility and reachability of objects w/r human Reachability of objects wiith respect to iself Computes a plan involving (potentially) both itself and the human Human action is selected when necessary Robot computes also grasps, placements and motions Robot executes plans and monitors H activity n n n By performing its own actions Asking human to achieve some actions Re-plans when necessary: state different from what is expected JOINT GOAL: CLEAN THE TABLE Kinect Camera + Pan/Tilt platfrom 12 Review Meeting Y1 26 13 Merci … 27 14