Whether it is improved resiliency in localization, robustness in control, or more efficient and faster learning of physical systems, the continuous enhancement of robotic systems has a long history of finding and exploring new mathematical structures tailored to the needs of these systems and their tasks. Symmetry, exploiting the tools of Lie group theory and differential geometry, has proven to be a powerful framework to analyse and improve real-world algorithms leading to significant performance gains across a range of fields such as state estimation, machine vision, AI, and control. The goal of this workshop is to provide a platform for leading researchers to advocate for this new perspective by highlighting and explaining the benefits of using symmetry as the foundation for algorithm analysis and design compared to classical methods. Viewing the world through the lens of symmetry, and exploiting invariance and equivariance properties, allows for a simpler and more holistic reexamination of classical problems.
The workshop will be structured in three parts:
In the first part, after an introductory tutorial on symmetry, two senior keynote speakers will deliver educational talks including state-of-the art applications on the use of equivariant systems theory in (a) state estimation and (b) control. This part will conclude with an intimate panel discussion between three senior experts in the field (disjoint from the keynote speakers). This change of protagonists on the stage will naturally foster discussions between the keynote speakers and the panelists, triggering also the audience's participation in a lively discussion.
The second part will feature young researchers presenting their recent findings in a poster session, sharing their journeys into this complex mathematical field.
The third part will consist of an educational third keynote talk on the benefit of equivariance in (c) artificial intelligence (of similar character to the first two keynote talks). A second, summarizing panel discussion with three senior experts and active audience participants will be moderated to merge the findings, questions, and challenges from all three subcommunities, thus highlighting novel avenues for the overall robotics community.
Both panel discussions will be guided by several overarching questions:
How can we provide an easy entry point from which to engage young scientists and other interested individuals in mathematically demanding areas of robotics, whether equivariant systems or others?
What defines a good symmetry for a given problem set, and in which fields do we expect new symmetries to have the greatest impact?
What are the open challenges in applying equivariant theory in robotics?
What can the equivariant systems community learn from other mathematically inclined areas of robotics that have already gone mainstream?
This workshop aims to foster a deeper understanding of the potential of equivariant system theory to enhance robotic systems and to educate and inspire the next generation of researchers.
Thank you all for joining us at RSS 2025 to explore these exciting developments!
Workshop Recordings
The YouTube playlist below includes recordings of all the background and keynote talks.
Speakers and Panelists
All speakers and panelists are confirmed for in-person attendance!
Panel Discussion: Kostas Daniilidis, Nadia Figueroa, David Rosen
17:00 - 17:45
Moderators: Maani Ghaffari, Jake Welde
Equivariance in Artificial Intelligence
Workshop Summary + Closing Remarks
17:45 - 18:00
Call for Contributions
We welcome submissions of extended abstracts to be presented (via spotlight talks and a poster session)
on-site on the workshop day June 25, 2025. The organizing committee will review and decide which submissions
will be presented at the workshop.
Apart from scientific excellence and relevance, the selection criteria
include the promotion of early-stage and student-driven research. The authors are encouraged to refine
their abstracts based on the received review and feedback gathered at the workshop, and ultimately submit
a full paper to the peer-reviewed venue of their choice (i.e. the workshop is non-archival). We also welcome
the submission of abstracts based on existing published/submitted work, but these submissions should clearly note the venue of publication.
Submission Guidelines
Submissions are welcome on the following non-exclusive list of topics:
Novel symmetries for sensors (navigation, manipulation, …)
Lie groups for estimation, control, and AI
Matrix Lie group methods
Symmetry-based approaches to trajectory tracking and control
Comparison of different symmetries highlighting their benefits and drawbacks in specific situations
Geometric deep learning (e.g., equivariant neural networks)
Invariance and canonicalization in machine learning
Symmetries for improved training phases as well as for improved inference
Limitations of symmetries in the fields of estimation, control, and AI
Modularity of symmetries
Mixed approaches using classical representations and symmetries
Handling broken or incomplete symmetry
Discrete symmetry groups
Variational integrators for mechanical systems
Connections between foundation models and symmetry
Real-world/simulated applications
Please submit your extended abstract as a PDF file via the OpenReview submission portal. Contributions should be no longer than 4+N pages in RSS double column format,
including figures and N pages of references. (The only content on the 5th and later pages should be references.)
Best Student Poster Award
A “Best Student Poster Award” with a cash prize of USD 250 will be given to the top poster with a student as the first author, sponsored by the organizers.
Important Dates
Deadlines are at 23:59 Anywhere on Earth (UTC-12).
Deadline
Notification
Early-Bird Submission
May 06, 2025
May 12, 2025
Last-Minute Submission
May 31, 2025
June 4, 2025
Submissions received after the Late-Breaking deadline may possibly be accepted at the discretion of the organizers (space permitting), although an in-depth review may not be possible. Please reach out over email with any concerns.
Note that after June 5, the late registration fees apply for the conference/workshops.
Accepted Contributions
Accepted extended abstracts are listed below - please attend the workshop to find out about these exciting new research developments!
Linear Memory SE(2) Invariant Attention,
Ethan Pronovost, Neha Boloor, Noureldin Hendy, Peter Schleede, Andres Morales, Nicholas Roy
🎉 🏆 WINNER OF THE BEST STUDENT POSTER AWARD! 🏆 🎉
Learning the RoPEs: Better 2D and 3D Position Encodings with STRING,
Connor Schenck, Isaac Reid, Mithun George Jacob, Alex Bewley, Joshua Ainslie, David Rendleman, Deepali Jain, Mohit Sharma, Kumar Avinava Dubey, Ayzaan Wahid, Sumeet Singh, René Wagner, Tianli Ding, Chuyuan Fu, Arunkumar Byravan, Jake Varley, Alexey A. Gritsenko, Matthias Minderer, Dmitry Kalashnikov, Jonathan Tompson, Vikas Sindhwani, Krzysztof Marcin Choromanski
