We're hiring multiple tenure-track faculty positions NAME at Michigan with special interests in Asst Prof level and marine robotics and autonomy. The search is broad, and we consider all applicants from different disciplines. Contact me if you have any questions.

Check out our new survey paper: Toward General-Purpose Robots via Foundation Models: A Survey and Meta-Analysis! arXiv: arxiv.org/abs/2312.08782 Website: robotics-fm-survey.github.io

[RAL+ICRA'24] A principled approch for underwater true color correction, learning 3D geometry and physics of light-water interaction together. code: github.com/tyz1030/neural… arxiv: arxiv.org/abs/2304.03384

We generate spectrograms that look like natural images by composing together the score functions of audio and visual diffusion networks. This was a fun project to be a part of! w/ Ziyang Chen and Daniel Geng Site: ificl.github.io/images-that-so… Paper: arxiv.org/abs/2405.12221

What happens when you train a video generation model to be conditioned on motion? Turns out you can perform "motion prompting," just like you might prompt an LLM! Doing so enables many different capabilities. Here’s a few examples – check out this thread 🧵 for more results!

We are excited to share our #NeurIPS2024 work "Multi-Object 3D Grounding with Dynamic Modules and Language Informed Spatial Attention" (D-LISA). Please drop by our poster session on Dec 11th 11 a.m. to 2 p.m. (Poster #1501) and learn from 楊喬諳 on Multi-Object 3D Grounding!

Command humanoids *directly* with natural language? Introducing LangWBC, a generative, end-to-end policy that turns natural language into real-world whole-body humanoid control! 💬→🦿Smooth, robust, surprisingly intuitive! See more 👉 LangWBC.github.io #RSS2025

the elegance of a single camera and unlimited field-of-view. I see a roboneck not just for humanoids, but also wheeled robots running on rugged terrain, subsea robots that cannot row and pitch, ….

Self-Questioning Language Models: LLMs that learn to generate their own questions and answers via asymmetric self-play RL. There is no external training data – the only input is a single prompt specifying the topic.