🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

DRIFT: Dead Reckoning In Field Time github.com/UMich-CURLY/dr… > Dead Reckoning In Field Time (DRIFT) is an open-source C++ software library designed to provide accurate and high-frequency proprioceptive state estimation for a variety of mobile robot architectures.

Guest lecture by Chien Erh Lin on Place Recognition covering old and new. Very educational and enjoyable to watch! Also, thanks to Kasra Khosoussi for sharing his slides a few years ago. youtube.com/watch?v=ueeeGs…

Grateful for the chance to give a TEDxMIT talk last week: Spatial AI for Robots and Humans youtube.com/watch?v=NGzuQj… - with thanks to my many wonderful collaborators, including Kurran Singh, Jungseok Hong, Ran Choi, Mike Benjamin, Ray Turrisi and Ge Yang

🤖 Excited for #RSS2024? Don’t miss our workshop on "Geometric and Algebraic Structure in Robot Learning"! Submit workshop papers (by 6/10 AOE) and dive into discussions on leveraging these structures for enhanced #robotics. 🚀🔍 Join us on 07/19 in Delft, Netherlands!

Every neuron is a Lie group (category). Lie Neurons (ICML'24), a conjugation equivariant neural network that models any semisimple Lie algebras. Same performance on training and test data in learning a 3D rotating rigid body from a single trajectory. arxiv.org/abs/2310.04521

We extended the workshop paper deadline to 06/17 AOE. Feel free to submit if you didn't catch! ✅ We also welcome volunteers for reviewing the workshop papers, even if you are an author! 🔖

Come stopping by our EquiVision workshop tomorrow at Summit 321 with the exciting talks!

The EquiVision workshop at #cvpr was amazing! Thank you Congyue Deng Yinshuang Xu Kostas Daniilidis Jiahui and all the organizers for making this happen!

I'll be presenting Lie Neurons at ICML Conference tomorrow (Wed) at Poster Session 4 (1:30 pm- 3:00 pm), Hall C #810. Come chat with me if you're also passionate about geometric deep learning or robotics! Paper: arxiv.org/abs/2310.04521 Code: github.com/UMich-CURLY/Li… #ICML2024

And we wrap GRaM Workshop at ICML 2024 with a social! Erik Bekkers Hannah Lawrence Oumar Kaba Alison Putri van der Linden Mona Schirmer Gabriele Cesa Julian Suk Guillermo Bernárdez Lev Telyatnikov and many many more!

In our recent IEEE RAS RAL paper, Chien Erh Lin & Minghan Zhu developed SE3ET: SE(3)-Equivariant Transformer for Low-Overlap Point Cloud Registration. Equivariant representation + Equivariant transformer -> Correspondences -> RANSAC pose registration.

I presented my RA-L paper SE3ET at #ICRA40. It was a unique experience, with incredible lineup of keynote speakers sharing their experience, debating various topics, and reflecting on the 40 years of ICRA.

Join us on Monday October. 14th at 2pm (UTC+4) in #IROS2024 Workshop on Equivariant Robotics. A great lineup of keynote speakers will discuss how symmetry penetrates each and every subfield of robotics. Website and Zoom Link: equirob2024.github.io

I am launching Duke Robot Dexterity Lab (DexLab) with a focus on human-level dexterity in robotic manipulation. The DexLab is looking for PhD students passionate about robotic manipulation starting Fall 2025!

I'm excited to share that I will defend my PhD dissertation, "Computational Symmetry and Learning for Generalizable Robot Perception," on Tuesday, December 3, at 3 PM EST! Zoom link and info:events.umich.edu/event/129461

Congrats to Tzu-Yuan Lin for successfully defending his PhD thesis! Justin made significant contributions to geometric robot state estimation & perception and geometric learning by developing the DRIFT library, Lie Neurons, with more ongoing efforts on perspective-equivariant

A big congrats to Tzu-Yuan Lin for defending a PhD last month: The work explores the incorporation of symmetric structures to develop efficient and generalizable perception algorithms for robotics. Watch: youtube.com/watch?v=iZPSY1…

Accepted to ICRA 2025! 🎉 We introduce MI-HGNN, a morphology-informed heterogeneous graph neural network built directly from a robot’s URDF. By embedding robot morphological constraints, MI-HGNN significantly enhances model efficiency and learning performance. Applied to contact

I graduated 6 PhD students since last year! Ray Zhang -> Research Scientist at Toyota Research Institute (TRI) Tzu-Yuan Lin -> Postdoc at MIT w/ Sangbae Kim Chien Erh Lin -> Applied Scientist at Amazon Robotics Tribhi Kathuria -> Lecturer at Michigan Robotics Sangli Teng -> Postdoc at UC

Thesis defended! Thank everyone for this wonderful 5-year journey at Stanford AI Lab Thank my advisor Leonidas Guibas for teaching me how to do research. The two most important words I learned from him are: passion and responsibility.