Launching a new humanoid learning course at @CMU_Robotics in Spring 2027!

Every team will get its own (small) humanoid robot for hands-on labs.

Will open-source everything (lecture slides, codes). Stay tuned!

How your robot moves depends on what it’s driving over, especially when moving at >10m/s over loose gravel or ice 🚙💨

What if dynamics models could be conditioned on exactly the terrain they expect to drive over? Introducing SpatialBEV!

Accepted at #CoRL2026 🤠
0/7

Common belief: Robot policies need continuous action heads; discrete actions are 𝐥𝐞𝐬𝐬 𝐧𝐚𝐭𝐢𝐯𝐞 𝐚𝐧𝐝 𝐥𝐞𝐬𝐬 𝐩𝐫𝐞𝐜𝐢𝐬𝐞.

❗We find policies w/ the simplest discrete representations 𝐦𝐚𝐭𝐜𝐡 𝐨𝐫 𝐛𝐞𝐚𝐭 continuous ones—in sim & real, in single-step & diffusios why 🧵

What to remember in the past to act—spatial, temporal or motion details?

We often assume what’s useful, but that’s hard to know in advance.

Meet IMPRINT 🧠: retain rich memory first, then learn to use it for control. Compact latents keep inference fast as history grows. 🧵 1/10

“As educators ourselves in a classroom, we need to adapt our strategy to make sure students understand and learn. That’s a critical capability in ARISTOS.”

Work could look completely different in ten years. @RamananDeva, @ChangliuL, Raj Reddy, Katia Sycara and @junyanz89 are platform for training tomorrow's skilled tradespeople. This summer, @ChrisRytting and team stopped by @CarnegieMellon to hear how advances in computer vision and video generation are enabling a tool that, like a good teacher, adapts as people learn.

More on the Moonshot: laude.org/moonshots/one/…

Placing a peg in a hole is child’s play — until it's 22,000 miles in space! 🛰️

@CMU_Robotics and @northropgrumman are teaching space robots how to dock, refuel, and fix old satellites, ending the days of single-use spacecraft: ri.cmu.edu/peg-in-hole-is…

Bonus pic: Sprout settling in at CMU with its very own RI @CMU_Robotics mug ☕️🤖

As part of #IROS2026, hundreds of academics and industry leaders toured CMU's robotics + engineering facilities on Oct. 1.

They got an inside look at what CMU researchers are working on inside the Robotics Innovation Center, Mill 19 and @NREC_CMU! 👀 🧠

bit.ly/4j9jLob

Robots just got a lot more delicate! 🤖🤏

@CMU_Robotics researchers designed human-inspired robotic fingertips with silicone skin & tiny fingernails. By combining force & sound sensors, the hands can handle fragile items without crushing them! #IROS2026

bit.ly/3U3L0Gq

I'm honored to be part of this amazing cohort of scientists! AI interacts with millions of people via screens, but the next step is interaction with the physical world via robots. Our @schmidtsciences project studies how robot "brains" can be safely deployed in the physical world

As part of Citadel founder and CEO Ken Griffin’s transformative investment, we’re now known as the Kenneth C. Griffin School of Computer Science! With this new support, we’re ready to ask bigger questions!
cmu.edu/news/stories/a…

Long-Horizon Adaptive Manipulation Planning (LAMP) helps robots plan not only where an object needs to go, but also how they can work together to get it there🤖🤝

MSR student and project co-lead Shuai Zhou presented this research at #IROS2026 yesterday!

ri.cmu.edu/lamp-helps-rob…

Humanoids can’t always just hit the brakes and power down 🤖🛑

CMU researchers and @Siemens developed PRISM, a new way to monitor when a humanoid can safely stop, helping robots know when to stop before things get risky! #IROS2026

ri.cmu.edu/knowing-their-…

1/ We introduce TrackEverything: a 3D point tracker that tracks all points across all frames of long videos (1000+ frames).

Our key idea is to tie computation to unique 3D scene content rather than redundant 2D pixels in a video.

trackeverything.github.io

Build the future of Physical AI with Fujitsu.
Fujitsu Research’s Physical AI team is excited to participate in #IROS2026 (September 27–October 1), bringing together researchers and innovators from around the world to advance intelligent robotics and embodied AI.

📅 September 27 hop with invited talks, spotlight sessions and awards. worlds-iros2026.github.io
📍 Booth #300 - Meet Fujitsu Research Physical AI team

We’re actively hiring researchers and engineers globally to advance Physical AI, robotics, embodied intelligence, robot learning, and spatial intelligence.

Interested in cutting-edge research, global collaboration, and turning research into real-world impact? Come talk with us about our work, including our collaboration with Carnegie Mellon University (CMU), and explore opportunities to join our growing team.

🌏 Open positions
🔴 Japan
✨ Embodied AI Research Scientist
jobs.global.fujitsu.com/job/Embodied-A…
✨フィジカルAIと空間知能化に関する研究 jobs.global.fujitsu.com/job/%E3%83%95%…
✨ 異種ロボットが協調動作するフィジカルAIの研究開発
jobs.global.fujitsu.com/job/%E7%95%B0%…

🔵 United States
✨ Physical AI Principal Researcher
jobs.global.fujitsu.com/job/Physical-A…
✨ Physical AI Senior Researcher
jobs.global.fujitsu.com/job/Physical-A…

🟢 India
✨ Senior Researcher (Physical AI and Robot Learning)
jobs.global.fujitsu.com/job/Senior-Res…

🤝 Meet us at IROS 2026, Booth #300. Whether you’re an experienced researcher, engineer, postdoctoral researcher, or emerging talent, come start a conversation about your future with Fujitsu Physical AI.

#PhysicalAI #Robotics #EmbodiedAI #IROS2026 #ResearchCareers @CMU_Robotics

Carnegie Mellon's robotics expertise will be on global display next week as thousands of researchers and industry leaders convene in Pittsburgh for #IROS2026!

See you there! 🤖 🤝

bit.ly/3V9mSSW

Manipulation is harder when the robot flies. 🚁

Moving the arm can shift the drone, and wind can push it off course. The policy and controller have to work together.

Here's AM-Bench: A Simulation Suite and Benchmark for Aerial Manipulation
ambench.github.io

#CoRL2026

Why do humans like to watch robots fall?

Associate Professor Changliu Liu (@ChangliuL) weighs in on this phenomenon on CNN! 👇 #TartanProud

cnn.com/2026/09/11/wor…

Humans learn by interacting with the world: How can we teach robots to similarly adapt to ambiguous dynamics? Introducing Wiggle and Go! #CoRL2026. We find that a specialized model can learn rope properties from just one interaction, and complete complex dynamic tasks.

The robot gures out how it behaves, then goes, using the estimated dynamics to execute the task. No real-world training data. No retries. #robotics #robotlearning

We introduce GTA-2, a multi-VLM framework that synthesizes executable robot manipulation skills from reusable, object-centric task-axis components

Zero-shot, no task-specific demonstrations or training.

gta2-project.github.io 🧵 1/9