Differences

This shows you the differences between two versions of the page.

--- news [2019/12/03 06:26]
IIS Webadmin
+++ news [2019/12/06 06:25]
IIS Webadmin
@@ Line 3: / Line 3: @@
 See our [[:|front page]] for the most recent news.
-<HTML><table><tr valign="top"><td>2019-10-10</td><td>IIS guest <span style="font-weight:bold">Georg Martius</span> gives an invited colloquium <i>Self-Organization of Behavior in Autonomous Robot
+<HTML><table><tr valign="top"><td>2019-10-10</td><td>IIS guest <span style="font-weight:bold">Georg Martius</span>, Max Planck Institute for Intelligent Systems,
-    Development</i> at IFI Lunchtime Seminar.</td></tr><tr valign="top"><td>2019-07-17</td><td>Matteo Saveriano contributes a talk <i>From Intuitive Skill Transfer to Large-Scale Robotic Knowledge Bases</i> at <a href="https://www.uibk.ac.at/disc/events/">DiSCourse - The Digital Science Seminar Series</a>, Department of Computer Science, University of Innsbruck.</td></tr><tr valign="top"><td>2019-07-15</td><td>David Peer contributes a talk <i>Vector Routing in Capsule Networks</i> at School of Learning, Perception and Robotics (SPROuT), Department of Information Engineering and Computer Science, University of Trento.</td></tr><tr valign="top"><td>2019-07-15</td><td>Sayantan Auddy contributes a talk <i>Progress, Compress and Expand - A framework for lifelong learning of
+    Tübingen,  gives an invited colloquium <i>Control What You Can – Intrinsically Motivated
+    Hierarchical RL</i> at IIS Colloquium. <span class="actions"><a href="javascript:void(0)" onclick="showHide('newsitem11', 'Abstract')">[Abstract]</a></span><blockquote id="newsitem11Abstract" style="display:none">I will present present a hierarchical reinforcement
+    learning agent that is intrinsically motivated to learn how to
+    control its observation space in the fastest possible manner by
+    optimizing learning progress. The agent mimicks to some extend the
+    development process how a baby learns.  Our agent learns what can
+    be controlled, how to allocate time and attention, and the
+    relations between objects in the environment.  In a nutshell, our
+    work combines several task-level planning ideas (e.g.,
+    backtracking search on task graph, probabilistic road-maps,
+    allocation of search efforts, etc.), in the form of structured
+    prior, with deep RL control and relational reasoning to learn from
+    scratch.</blockquote></td></tr><tr valign="top"><td>2019-10-10</td><td>IIS guest <span style="font-weight:bold">Georg Martius</span>, Max Planck Institute for Intelligent Systems,
+    Tübingen,  gives an invited colloquium <i>Self-Organization of Behavior in Autonomous Robot
+    Development</i> at <a href="https://www.uibk.ac.at/informatik/forschung/lunchtime-seminar/index.html.en">IFI Lunchtime Seminar</a>. <span class="actions"><a href="javascript:void(0)" onclick="showHide('newsitem12', 'Abstract')">[Abstract]</a></span><blockquote id="newsitem12Abstract" style="display:none">I am studying the question how robots can autonomously
+    develop skills.  Considering children, it seems natural that they
+    have their own agenda. They explore their environment in a playful
+    way, without the necessity for somebody to tell them what to do
+    next. With robots the situation is different. There are many
+    methods to let robots learn to do something, but it is always
+    about learning to do a specific task from a supervision
+    signal. Unfortunately, these methods do not scale well to systems
+    with many degrees of freedom, except a good prestructuring is
+    available.  The hypothesis is that if the robots first learn to
+    use their bodies and interact with the environment in a playful
+    way they can acquire many small skills with which they can later
+    solve complicated tasks much quicker. In the talk I will present
+    my steps into this direction. Starting from some general
+    information theoretic consideration we provide robots with an own
+    drive to do something and explore their behavioral
+    capabilities. Technically, this is achieved by considering the
+    sensorimotor loop as a dynamical system, whose parameters are
+    adapted online according to a gradient ascent on an approximated
+    information quantity.  I'll show examples of simulated and real
+    robots behaving in a self-determined way and present future
+    directions.</blockquote></td></tr><tr valign="top"><td>2019-07-17</td><td>Matteo Saveriano contributes a talk <i>From Intuitive Skill Transfer to Large-Scale Robotic Knowledge Bases</i> at <a href="https://www.uibk.ac.at/disc/events/">DiSCourse - The Digital Science Seminar Series</a>, Department of Computer Science, University of Innsbruck.</td></tr><tr valign="top"><td>2019-07-15</td><td>David Peer contributes a talk <i>Vector Routing in Capsule Networks</i> at School of Learning, Perception and Robotics (SPROuT), Department of Information Engineering and Computer Science, University of Trento.</td></tr><tr valign="top"><td>2019-07-15</td><td>Sayantan Auddy contributes a talk <i>Progress, Compress and Expand - A framework for lifelong learning of
 robotics tasks</i> at School of Learning, Perception and Robotics (SPROuT), Department of Information Engineering and Computer Science, University of Trento.</td></tr><tr valign="top"><td>2019-07-15</td><td>Justus Piater contributes a talk <i>Towards Open-Ended Robot Learning</i> at School of Learning, Perception and Robotics (SPROuT), Department of Information Engineering and Computer Science, University of Trento.</td></tr><tr valign="top"><td>2019-07-15</td><td>Matteo Saveriano contributes a talk <i>Learning Structured Robotic Tasks via Human Imitation</i> at School of Learning, Perception and Robotics (SPROuT), Department of Information Engineering and Computer Science, University of Trento.</td></tr><tr valign="top"><td>2019-07-12</td><td>Simon Haller and Nicolas Stolz appear in the media: Sport und Roboter gehen Hand in Hand (in Liechtensteiner Vaterland).</td></tr><tr valign="top"><td>2019-07-04</td><td>Simon Haller teaches a tutorial <i>Introduction to Prototyping autonomous Robots using the Robot
 Operating System (ROS)</i> at Institut für Informatik, Universität Innsbruck.</td></tr><tr valign="top"><td>2019-06-27</td><td>Justus Piater gives an invited talk <i>Selbst lernende Roboter der Zukunft</i> at <a href="http://www.gmar.at/aktuell/">GMAR Robotics Talks
@@ Line 16: / Line 51: @@
     Workshop on Computational Models of Affordance in Robotics
     </a>.</td></tr><tr valign="top"><td>2019-05-09</td><td>Erwan Renaudo contributes a talk <i>Action representations in robotics: A taxonomy
-    and systematic classification</i> at <a href="https://www.uibk.ac.at/informatik/forschung/lunchtime-seminar/events/2019-05-09.html">
+    and systematic classification</i> at <a href="https://www.uibk.ac.at/informatik/forschung/lunchtime-seminar/events/2019-05-09.html">IFI
     Lunchtime Seminar</a>, Universität Innsbruck.</td></tr><tr valign="top"><td>2019-05-02</td><td>Justus Piater gives an invited talk <i>Digitalization: Promises and Challenges</i> at <a href="https://www.andrassyuni.eu/nachrichten/innovation-trifft-geschichte-oder-wie-kommunizieren-wir-digital-und-analog.html">Möglichkeiten
     und Herausforderungen der Digitalisierung</a>, Universität Innsbruck.</td></tr><tr valign="top"><td>2019-04-26 – 2019-04-27</td><td>Birgit Juen, Cornelia Vidovic, Manuela Resch, Justus Piater, and Simon Haller co-organize the <a href="https://www.uibk.ac.at/informatik/schule/robocupjunior/index.html.de">RoboCup Junior Austrian Open 2019</a>.</td></tr><tr valign="top"><td>2019-04-26</td><td>Justus Piater appears in the media: TV interview on educational and societal aspects of robot

User Tools

Site Tools

Differences

Page Tools