Monday, July 15, 2024 (Auditorium Aula-20, Aula Conference Center, TU Delft, Netherlands)
Workshop on Embodiment-Aware Robot Learning (EARL)
at Robotics: Science and Systems (RSS)
This workshop brings together researchers working on co-design of robot embodiment and control algorithms
Speakers
Stanford & Columbia University
Vrije Universiteit Amsterdam
Stanford University
Boston Dynamics
Columbia University
CSIRO Australia
University of Vermont
Venue
EARL Workshop at RSS'24
TU Delft, Netherlands
The workshop on Embodiment-Aware Robot Learning (EARL) brings the optimization of robot embodiment into the foreground of robot learning. Artificial Intelligence (AI) in robotics has come to primarily mean intelligent decision making within a fixed embodiment. However, natural intelligence is broader, and it includes optimization of the organism’s body for its ecological niche. We, therefore, seek to emphasize the importance of co-design of the agent’s body and behavior within the field of robot learning and AI, both from a conceptual point of view and in practical terms.
We welcome participants from various sub-communities, working on both hardware and software aspects of robotics, such as bioinspired robotics, field robotics, grasping and manipulation, robot learning, control & dynamics, mechanisms & design, and robot modeling & simulation, among others.
Call for Papers
We invite extended abstracts (2-4 pages excluding references, in RSS paper format) to be presented as lightning talks and posters during the workshop. We welcome submissions that report new results, ongoing research, and broad vision papers. All submissions are double-blind and peer-reviewed. We invite participants from a variety of backgrounds: robotics, machine learning, evolutionary optimization, mechanical design, electrical engineering.
Submission Deadline (Extended): June 14
Acceptance Notification: June 24
Camera-ready Paper Submission Deadline: July 7
Best poster award:
Carmelo Sferrazza, Dun-Ming Huang, Fangchen Liu, Jongmin Lee, Pieter Abbeel
Body Transformer: Leveraging Robot Embodiment for Policy Learning
Best paper award:
Sergio Hernández-Gutiérrez, Ville Kyrki, Kevin Sebastian Luck
Following Ancestral Footsteps: Co-Designing Agent Morphology and Behaviour with Self-Imitation Learning
Schedule
08:50-09:00
Opening Remarks
09:00 - 09:30
Environment and Embodiment-Aware Design Optimisation
David Howard, Commonwealth Scientific and Industrial Research Organisation
09:30-10:00
A Full Stack Approach to Dexterous Manipulation
Pulkit Agrawal, Massachusetts Institute of Technology
Morning Coffee Break 10:00 - 10:30
10:30-11:00
Co-Adaptation of Robot Design & Behavior: A Reinforcement Learning Perspective
Kevin Sebastian Luck, Vrije Universiteit Amsterdam
11:00 - 11:30
Enabling Cross-Embodiment Policy Learning
Jeannette Bohg, Stanford University
11:30 - 12:00
Hardware and Policy Co-design with Deep Reinforcement Learning for Robotic Manipulators
Matei Ciocarlie, Columbia University
12:00 - 12:30
Lightning Talks by Workshop Participants
Lunch Break 12:30 - 14:00
14:00-14:30
Self-Adaptive Robots
Shuran Song, Stanford University
14:30 - 15:00
Structural Priors for Robot Learning
Michael Lutter, Boston Dynamics
15:00 - 15:30
On the Fragile Co-Adaptation of Brain and Body in Evolved Soft Robots
Nick Cheney, University of Vermont
Afternoon Coffee Break 15:30 - 16:00
16:00 - 17:00
Poster Session
17:00 - 17:15
