Title: Motion generation for humanoid robots: machine learning and industrial
Prof. Olivier STASSE, senior CNRS researcher at LAAS, Toulouse in Gepetto group
Abstract. In this talk, I will present the scientific approach followed by the Gepetto team to tackle the problem of motion generation for humanoid robots. In the context of industrial application involving interaction with the environment, our current motion generation pipe-line based on optimization, control and specific structure of mechanical systems will be outlined. In this rapidly changing environment I will present scientific challenges lying ahead where machine learning together with mechatronics and optimization are playing pivotal roles.
Bio. O. Stasse is a CNRS Senior Researcher at LAAS, CNRS. His research topics are on humanoid motion generation, perception for humanoids, fast planning and reactive control. He received a Ph.D. in Intelligent Systems from the University of Paris 6 in 2000 under the supervision of P. Coiffet. From 2003 to 2011, he was at the Joint French-Japanese Robotics Laboratory (JRL) CNRS/AIST in Tsukuba, Japan. In 2011 he joined the Gepetto team at LAAS, Toulouse. He is an associate editor of RA-L since its foundation. An IEEE Senior Member, he was nominated as an exceptional reviewer in 2012 for the IEEE Transactions on Robotics journal in 2012. Together with Bjorn Verrelst he won the best paper Award at ICMA 2006 for stepping over large obstacle with the humanoid robot HRP-2. Along with Francois Saidi he was finalist for best paper award ICAR 2007 for visual search with HRP-2. Together with Andrew Davison, Nicolas Mansard and other colleagues he was finalist for best video award ICRA 2007. He recently specified the new humanoid platform for the Gepetto group. The robot was realized by PAL Robotics and is now a product.
Engineering ethical large-scale multi-agent systems
Prof. Yves DEMAZEAU, CNRS Research Fellow, Laboratoire d’Informatique de Grenoble
Abstract. Engineering complex multi-agent systems (MAS) requires to evolve towards more scalable and ethical systems. We propose an iterative method of constructing such scalable and ethical systems, divided into six stages, each of them being independently verified, the whole system being empirically validated. We investigate how to determine, at design time, the right balance between legal aspects and moral aspects of the agents involved, and how this balance evolves over time. We illustrate the methods in several domains including unmanned aerial vehicles.
Bio. Yves Demazeau received his PhD in Computer Science at INP Grenoble Institute in 1986. He is Director of Research at CNRS since 1988, and Invited Professor at SDU since 2002. He has been distinguished ECCAI Fellow in 2010 and Doctor Honoris Causa at University of Southern Denmark Odense in 2017. Yves Demazeau is one of the founders of the Multi-Agent Systems (MAS) area. His research focuses on engineering multi-agent systems. He has edited 39 books or proceedings, has authored or co-authored 190 papers and has been invited for 54 lectures, 22 tutorials, and 196 seminars. He has (co-)advised 57 MSc students and 29 PhD students and served in 278 conference program committees. He is conducting the PAAMS series since 2009 and the National AI Platform PFIA since 2013. He served as Editor-in-Chief of the French AI Journal (RIA) from 2010 to 2018. He is the President of the French Association for Artificial Intelligence (AFIA) since 2011.