Delft Workshop on Robot Learning (DWRL) 2017

A Delft Robotics Institute sponsored Interdisciplinary Thematic Workshop

We have a few remaining spots available, for which you may join without submitting an abstract. If you want to attend, please inform us by sending an email to dwrl2017@gmail.com.

 

Workshop date: October 2nd, 2017.

Submission deadline: September 20th, 2017.

Location: Mekelzaal 4 (MSB4), Science Centre Delft, Delft University of Technology. 
Mijnbouwstraat 120, 2628 RX Delft, The Netherlands.

Organizers: Joost Broekens (TU Delft), Jens Kober (TUD), Matthijs Spaan (TUD), Thomas Moerland (TUD).

Invited speakers: Guszti Eiben (Vrije Universiteit Amsterdam), Sven Behnke (Universitat Bonn), Sylvain Calinon (Idiap Research Institute), Frans Oliehoek (University of Liverpool).

Contact: dwrl2017@gmail.com

 

Abstract
The aim of this one-day workshop is to gather scientist that are working on the intersection of machine learning and robotics. Applications of machine learning techniques in the robotic setting are strongly increasing, e.g., in movement generation, in decision making, in control, in vision, in perception, in design optimization, etc. At the same time, high-impact machine learning publications seem to put less-and-less emphasis at data efficiency. Thereby, these implementations are hard to transfer to real-world robots, which can only learn in wall-clock time and suffer from wear-and-tear. The goal of this workshop is to identify the common themes and challenges in learning techniques for robots, with researchers from a variety of robot learning directions. We also encourage participation from non-robotic machine learners (i.e. with work in simulation), when they are interested in joining the discussion the long-term horizon for their techniques in real-world applications.

 

Topics
Reinforcement learning, deep learning for robotics (e.g., in object recognition, task learning), representation learning, planning under uncertainty, imitation learning, programming by demonstration, adaptation and adaptive behaviour,  evolutionary robotics.

 

Format
One day workshop with approximately 30 participants. All participants submit a 2-page abstract in advance, and present a poster and a 5 minute flash talk. 4 invited speakers. Two discussion slots. Lunch and coffee included, optional dinner at own cost.