SAIR Lab

🚀 VL-Nav = language → logic → action VL-Nav brings neuro-symbolic intelligence to vision-language navigation 🎥 Watch more demos at sairlab.org/vlnav/

Cheers! This marks a solid step in our lab forward for vision-language navigation on mobile robots. More demos will be released soon.

Navigating a mobile manipulator among movable obstacles is tough. We built a fast task planner using neuro-symbolic relaxation to boost both success rate and planning speed. Successfully verified on a quadruped manipulator! Paper accepted to RA-L. Details: sairlab.org/flax/

It has been my great pleasure to help make this happen. Looking forward to deepening insight and impact in 2026. Check out more photos and milestones at sairlab.org/2025/

🎉 Happy New Year 2026! 🚀 We’re excited to share that PyPose surpassed 160,000+ pip downloads in 2025 alone, representing over 60× growth since its creation in 2022. This indicates a strong and growing popularity trend. Grateful for the support, and excited for what’s ahead! ✨

As we wrap up 2025, we’re excited to share a summary of SAIR Lab’s key milestones from this year: 👉 sairlab.org/2025/ Looking ahead, our theme for 2026 is 👉 Insight 👈, shaping a clearer, more ambitious long-term direction for robot autonomy research. 🚀

🚀 Thinking about applying for a PhD in Robotics? Check out roboranking.org, a website where you can quickly match universities and faculties. 💡 Hint: Don’t just look at rankings — focus on your research interests first!

🚨 Happy to share "iWalker: Imperative Visual Planning for Walking Humanoid Robot" accepted to IROS'25! It’s a small step for a humanoid robot, but a giant leap for imperative learning (IL) in legged robotics! Project: sairlab.org/iwalker/ IL articles: sairlab.org/iseries/

Great to see Neuro-symbolic AI gaining momentum in computer vision, especially in the context of robotics and embodied AI problems. Inspiring talk by @DrChenWang at the @CVPR AC Workshop! #CVPR2025 #EmbodiedAI #NeuroSymbolic #ComputerVision #AIResearch

We're happy to share that Imperative Learning has been formally accepted to The International Journal of Robotics Research (IJRR) — one of the most prestigious journals in robotics! Please refer to our project page for more details: sairlab.org/iseries/

Excited to introduce our latest research "VL-Nav: A Real-Time Vision-Language Navigation System with Spatial Reasoning". It enables tasks like "find a person wearing black" and runs 30 Hz on a Jetson Orin NX! Check more details at sairlab.org/vlnav/

We are happy to announce that the DDL for the ICRA'25 workshop call for papers on "Foundation Models and Neuro-symbolic AI for Robotics” have been extended to April 1 (no more extension). More information about it can be found at sairlab.org/icra25/

By integrating hundreds of users’ feedback and adding data for 2025, we are now able to provide a more comprehensive Roboranking, the robotics faculty & the university ranking system. Check here to see the latest version of roboranking.org

🎉 Exciting News! 🎉 Thrilled to share that our latest V-SLAM system, AirSLAM, has been officially accepted to IEEE Transactions on Robotics (TRO)! 🚀 We’re proud to make this open-source for the community—feel free to download, experiment, and innovate! 🚀sairlab.org/airslam/

As we bid farewell to 2024 and enter 2025, we’re excited to share a summary of SAIR Lab’s key milestones from this past year: sairlab.org/2024/. Looking ahead, our theme for 2025 is 👉 Transform 👈. A heartfelt thank you to all our team members! 🎉

🎉🎉🎉 I’m honored to contribute as one of the chapter authors with an incredible team for this incredible SLAM Handbook! This is just the beginning, and your feedback will help make it even better! 🙌

🎉 Success! We've applied Imperative Learning (IL) to Humanoid Robots for footstep planning! Next time let's walk a humanoid robot under diverse terrain! Check out the latest preprint at sairlab.org/iwalker/ @Tymonmeng @DrChenWang #Robotics #AI #HumanoidRobots

🚀 Exciting news! Check out our latest work on the Multiple Traveling Salesman Problem (iMTSP) in IROS'24! We're pushing the boundaries of efficient and scalable AI-driven scheduling solutions. Explore its source codes and paper at sairlab.org/iMTSP/ @DrChenWang

👉 Check out the code: github.com/sair-lab/PhysO… 🔗 Learn more: sairlab.org/physord/ #AI #Robotics #OffRoad

🚀 Excited to present PhysORD at #IROS2024! It is a neuro-symbolic model for motion prediction in off-road driving, enforcing the physical law of conservation into ML. PhysORD outperforms in accuracy, efficiency, and long-term generalization! arxiv.org/abs/2404.01596