- We will be collecting questions ahead of time (in addition to questions you can ask in real-time during panels).
- Please add your questions here.
- The collected questions will be discussed during the live panel sessions.
Aim and scope
Classical Imitation Learning methods sequentially obtain expert demonstrations or observations, which then are used to learn either policies or objective functions for a robot, and sometimes refined by a subsequent step of Reinforcement Learning. However, recent advances in Robot Learning have focused on learning loops that include human teachers participating in iterative, interactive, and recursive approaches. Here the learner receives intermittent signals of either preferences, reinforcements or corrective demonstrations. These signals are used to incrementally improve the policies. These improved policies in turn drive the agent to visit more relevant regions of the state space, resulting in more meaningful data and continued policy improvement.
Interactive Robot Learning methods are characterized mainly by the type of human feedback, but also cover a whole range of techniques combining Active Learning and/or (Inverse) Reinforcement Learning, in various settings such as on-line/off-line learning, model-free/model-based learning.
The aim of this workshop is to bring together the different perspectives of current developments in Interactive Robot Learning, which are going to lead the future generation of applications with presence of easily adaptable and flexible robots in smart industries, along with assistive functionalities in care and household environments. Easy to program robots are of interest for the ongoing industry revolution. Such new methods will allow non-expert robot users and non-expert task demonstrators to teach robots which attain superhuman performances.
We want to have the perspectives of renowned expert researchers in the field, and offer the young researchers and practitioners the opportunity to share and discuss their new ideas and preliminary results of ongoing research, in order to strengthen the path towards more efficient and effective methods and practices in the field.
The invited talks will be recorded and posted on this website some days before the event date. There will be a public videoconference for a panel discussion with the invited speakers during the event.
This workshop is supported by the following IEEE/RAS technical committees:
Topics of interest
- Learning from Demonstrations
- Active Learning / Active Teaching
- Interactive Reinforcement Learning
- Learning from Preferences
- Learning from Corrective Demonstrations
- Learning from Physical Interaction
- Interactive Inverse Reinforcement Learning
- Learning objectives for motion planning
- Model-free and model-based Imitation Learning
- Evaluation of Interactive Learners
- Curriculum Learning
program and Registration
Workshop Slack channel: #ws24
The workshop will be held with an asynchronous and a synchronous phase:
- Asynchronous phase, in which the participants will be able to watch the videos of the recorded talks of the invited speakers, along with the short presentations of the authors of the accepted contributions. This can be done at any time at the convenience of the participants. The videos will be posted on this website. It is recommended to watch the videos before the synchronous online meeting.
- Synchronous phase, for having a mixture of Q&A session and panel discussion. This online session will be held on June 5 at 15:00 UTC.
- 15:00-16:00 Panel with the authors of the contributions.
- 16:00-18:00 Panel with the invited speakers.
The registration process could be done with this form.
Please submit any questions to: firstname.lastname@example.org.
Call for Papers
In order to increase the participation of young and senior researchers in the topic, along with the quality of the contributions, we invite participants to contribute with papers and extended abstracts for two kind of submissions:
- Traditional submissions (paper or abstracts) presenting preliminary results of ongoing research.
- Submissions (abstract) linked to already published papers coming from:
- Journal papers, which have not had the chance to be discussed in a conference, (imitating the strategy of RA-L and TR-O).
- Conference papers, which come from Machine Learning conferences. Although these papers already had exposure in a conference, it is a great opportunity to give visibility to the most recent contributions, within a more specific robotics research community.
We encourage participants to submit their research in form of either a paper (6 pages maximum without references), or an extended abstract (2 pages maximum), both using the IEEE conference proceedings template. Contributors are encouraged to submit their manuscripts with videos illustrating their work.
The submitted contributions will go through a single blind review process. All the accepted contributions will have a spotlight presentation to be posted on our website before the event date, and there will be a session of Q&A with a public video conference moderated by the organizers during the event.
All submissions should be sent in PDF format to the email: email@example.com
Important Dates: (updated)
- Submission deadline: ̶A̶p̶r̶i̶l̶ ̶ ̶0̶3̶ May 18, 2020
- Notification of acceptance: ̶A̶p̶r̶i̶l̶ ̶1̶7̶ May 25, 2020
- Workshop date: ̶M̶a̶y̶ ̶3̶1̶ June 5th, 2020
- Carlos Celemin, TU Delft <C.E.CeleminPaez@tudelft.nl>
- Harish Ravichandar, Georgia Institute of Technology <firstname.lastname@example.org>
- João Silvério, Idiap Research Institute <email@example.com>
- Jens Kober, TU Delft <J.Kober@tudelft.nl>
- Maya Cakmak, University of Washington, <firstname.lastname@example.org>
- Sonia H Chernova, Georgia Institute of Technology <email@example.com>
- Sylvain Calinon, Idiap Research Institute <firstname.lastname@example.org>