Prof. Robert Babuska
Title: Data-driven construction of parsimonious analytic models for autonomous robots
Abstract:Developing mathematical models of robots and their environment is essential for long-term robot autonomy. Even popular model-free control techniques such as reinforcement learning benefit from the use of models, which are typically learned from data collected by the robot during its operation. In this talk, we introduce a family of methods that employ symbolic regression to construct parsimonious models described by analytic equations. Contrary to other techniques, such as deep neural networks, symbolic regression generates compact input-output or state-space models with small number of tunable parameters and with accuracy exceeding most competitive methods. In the context of long-term autonomy, we also address the problem of selecting informative data for model updating and we show examples of robots with up to 14-dimensional state space, including a wheeled mobile robot and a bipedal walker.
Bio:Prof. Robert Babuska received the M.Sc. (Hons.) degree in control engineering from the Czech Technical University in Prague, in 1990, and the Ph.D. (cum laude) degree from TU Delft, the Netherlands, in 1997. He has had faculty appointments with the Czech Technical University in Prague and with the Electrical Engineering Faculty, TU Delft. Currently, he is a full professor of Intelligent Control and Robotics at TU Delft, Faculty 3mE, Department of Cognitive Robotics.
In the past, he made seminal contributions also to the field of nonlinear control and identification with the use of fuzzy modeling techniques. His current research interests include reinforcement learning, adaptive and learning robot control, nonlinear system identification and state-estimation. He has been involved in the applications of these techniques in various fields, ranging from process control to robotics and aerospace.
Prof. Josef Sivic
Title: Visual recognition: from Internet images towards robots that see
Building machines that can automatically understand complex visual inputs is one of the central problems in artificial intelligence. In this talk, I will argue that in order to build machines that understand the changing visual world around us the next challenges lie in developing visual representations that generalize to never seen before conditions and are learnable in a weakly supervised manner, i.e. from noisy and only partially annotated data. I will show examples of our work in this direction with applications in visual localization across changing conditions or learning how people manipulate objects from instructional videos.
Josef Sivic holds a senior researcher position at Inria in Paris and a Distinguished Senior Researcher position at the Czech Institute of Robotics, Informatics and Cybernetics at the Czech Technical University in Prague where he leads a newly created team on Intelligent Machine Perception. He received the habilitation degree from Ecole Normale Superieure in Paris in 2014, PhD from the University of Oxford in 2006 and MSc degree from Czech Technical University in 2002. Before joining Inria he was a post-doctoral associate at the Computer Science and Artificial Intelligence Lab at the Massachusetts Institute of Technology. He received the British Machine Vision Association Sullivan Thesis Prize and his papers have been awarded the Longuet-Higgins prize (CVPR’07) and the Helmholtz prize (ICCV’03 and ICCV’05) for fundamental contributions to computer vision that withstood the test of time. He is a fellow of the Learning in Machines & Brains program at the Canadian Institute of Advanced Research and was awarded an ERC Starting Grant in 2013.