ICRA'25 Workshop on Foundation Models and Neuro-Symbolic AI for Robotics

The field of robotics has experienced significant growth through the integration of machine learning models, particularly with the rise of foundation models, neuro-symbolic AI, generative models (such as large language models and diffusion models), and other emerging techniques. While these methods have transformed areas like computer vision, natural language processing, and computer graphics, their application in robotics remains largely underexplored.

This workshop will address this emerging area by bringing together leading researchers and practitioners in robotics to share insights on the latest advancements, methodologies, and best practices. It will feature invited talks from academic and industrial leaders and technical paper presentations discussing groundbreaking approaches and robotic systems that leverage foundation models and neuro-symbolic AI. The workshop will also explore current challenges, future research directions, and promote interaction through poster sessions and spotlight talks.

Call for Workshop Papers

We invite submissions for this upcoming workshop, focusing on foundation models, neuro-symbolic AI, generative models, and their applications for robotics. The covered topics strictly align with the recent IJRR Special Issue on “Foundation Models and Neuro-Symbolic AI for Robotics”.

Submission site opens on Feb 1, 2025, closes on Apr 3, 2025 (no more extension); Notification on Apr 17, 2025; Camera-ready on Apr 28, 2025; Pacific Time (PT).
Format: 4 pages excluding acknowledgement, references, and unlimited appendix. Please use the ICRA’25 main conference template (IEEE conference style).

Our primary goal is to foster meaningful discussions and idea exchanges between workshop paper authors and the invited speakers. Therefore, we only accept 4-page short papers, which will not be considered formal publications. Authors may therefore either submit their summarized recent published/accepted results (E.g., IJRR, TRO, and venues selected by roboranking.org) or new ideas with preliminary and unpublished results. This policy allows authors to further develop or resubmit their work to other top-tier venues. Accepted workshop papers will be archived at OpenReview and will be promoted to a broader audience through various media channels.

Please note that we use OpenReview for paper reviewing, but reviewer comments will remain confidential, similar to a traditional review process. The reviewing will be single blind. All authors of accepted workshop papers will be required to present their work in a poster session. Authors of outstanding submissions will be invited to deliver an oral presentation during the seminar listed below. Best paper awards will be presented by the IEEE RAS Technical Committee (TC) on Computer & Robot Vision (CRV). To be eligible for these awards, authors must join the IEEE RAS TC on CRV (which is free) through the IEEE Membership Portal.

May 19, 2025, 8:30am - 5:30pm (Meeting Room: 305)

	Presenter	Session Title & Slides Link	Time
Awards Oral Presentation Opening: Chen Wang (3 min)		Award Candidate Talk 1 (5 min talk + 1 min Q&A): Tool-as-Interface: Learning Robot Tool Use from Human Play through Imitation Learning Award Candidate Talk 2 (5 min talk + 1 min Q&A): MAGIC-VFM: Meta-learning Adaptation for Ground Interaction Control with Visual Foundation Models	8:30 - 8:45 AM
	Luca Carlone Associate Professor Massachusetts Institute of Technology	Neuro-Symbolic Reasoning for 3D Spatial AI	8:45 - 9:05 AM
	Danfei Xu Assistant Professor Georgia Institute of Technology	Neuro-Symbolic Data Engine for Robot Learning	9:10 - 9:30 AM
	Jiayuan Mao PhD candidate Massachusetts Institute of Technology	Compositional Action Abstractions	9:35 - 9:55 AM
	Yunzhu Li Assistant Professor Columbia University	From Language to Code: Foundation Models for Task Specification in Robotic Manipulation	10:00 - 10:20 AM
Poster Session & Award Announcement Award Presenter: Luca Carlone Note: Award announcement at the beginning of the poster session (5 min)			10:25 - 11:00 AM
	Yisong Yue Professor California Institute of Technology	Neuro-symbolic AI for Safety-Critical Agile Control	11:00 - 11:20 AM
	Chen Wang Assistant Professor University at Buffalo	Self-supervised Neuro-Symbolic Learning for Robot Autonomy	11:25 - 11:45 AM
	Yilun Du Assistant Professor Harvard University	Generalizable Foundation Models through Compositional Generation	11:50 AM - 12:10 PM
	Luigi Palmieri Research Scientist Bosch Research	Context-aware Robot Navigation in Human-shared Environments	12:15 - 12:35 PM
Lunch Break			12:40 - 1:40 PM
	Ayoung Kim Associate Professor Seoul National University	Neuro-Symbolic Approaches for LiDAR Place Recognition and Transparent Object Reconstruction	1:40 - 2:00 PM
	Dinesh Jayaraman Assistant Professor University of Pennsylvania	Tapping Foundation Models for Rewards, Representations, and Policies	2:05 - 2:25 PM
	Anirudha Majumdar Associate Professor Princeton University	Foundation model-powered robots that know when they don't know	2:30 - 2:50 PM
	Nima Fazeli Assistant Professor University of Michigan	Dexterous Robots in the Age of Foundation Model: Tactile, Language, and Gestures Representations	2:55 - 3:15 PM
Poster Session & Social Networking			3:20 - 3:50 PM
	Sebastian Scherer Associate Research Professor Carnegie Mellon University	Combining neural representations with rule-based reasoning for autonomous drone search	3:50 - 4:10 PM
	Lu Gan Assistant Professor Georgia Institute of Technology	Morphology-Informed Neural Networks for Legged Robot Autonomy	4:15 - 4:35 PM
	Wenlong Huang Ph.D. student Stanford University	Generalization through Task Representations with Foundation Models	4:40 - 5:00 PM
	Jiajun Wu Assistant Professor Stanford University	Structured Representations in Actions, Objects, and Tasks for Robotics	5:05 - 5:25 PM