Ruihan Yang
@RchalYang
PhD from @UCSanDiego Robot Learning / Embodied AI
At IROS 2024 now. I’ll present our work HarmonicMM tomorrow 10AM at WeAT2 Also open to all kinds of discussion! Let me know if you’d like to chat!
How to tackle complex household tasks like opening doors, and cleaning tables in real? Introducing HarmonicMM: Our latest model seamlessly combines navigation and manipulation, enabling robots to tackle household tasks using only RGB visual observation and robot proprioception.
When it comes to scaling data, it’s not just about scale—it’s also about distribution. Leveraging generative models, even simple ones, can help improve both. Great work led by @jianglong_ye & @kaylee_keyi!
How to generate billion-scale manipulation demonstrations easily? Let us leverage generative models! 🤖✨ We introduce Dex1B, a framework that generates 1 BILLION diverse dexterous hand demonstrations for both grasping 🖐️and articulation 💻 tasks using a simple C-VAE model.
Our robotics team will be at ICRA next week in Atlanta! Having started a new research team at Amazon building robot foundation models, we're hiring across all levels, full-time or intern, and across both SW and Research roles. Ping me at [email protected] and let's have a chat!
that means you are so lucky to deeply understand multiple problems during your phd.
ummm… As a robotics PhD student, I’m genuinely worried that the problem I find important now will be solved in the next 2 years—by MORE DATA, without any need to understand the underlying structure. And this happens in many areas😂
Thank you @xiaolonw for all the support and guidance over the past six years! It’s been a truly transformative experience, and I’m so grateful for everything I’ve learned along the way. Hard to believe this chapter is coming to a close.
Congratulations to the graduation of @Jerry_XU_Jiarui @JitengMu @RchalYang @YinboChen ! I am excited for their future journeys in industry: Jiarui -> OpenAI Jiteng -> Adobe Ruihan -> Amazon Yinbo -> OpenAI
For years, I’ve been tuning parameters for robot designs and controllers on specific tasks. Now we can automate this on dataset-scale. Introducing Co-Design of Soft Gripper with Neural Physics - a soft gripper trained in simulation to deform while handling load.
AMO, a new humanoid controller, makes the humanoid not just move around with fancy motions, but actually do something useful. We will present this #RSS2025
Meet 𝐀𝐌𝐎 — our universal whole‑body controller that unleashes the 𝐟𝐮𝐥𝐥 kinematic workspace of humanoid robots to the physical world. AMO is a single policy trained with RL + Hybrid Mocap & Trajectory‑Opt. Accepted to #RSS2025. Try our open models & more 👉…
Amazing results.
π-0.5 is here, and it can generalize to new homes! Some fun experiments with my colleagues at @physical_int, introducing π-0.5 (“pi oh five”). Our new VLA can put dishes in the sink, clean up spills and do all this in homes that it was not trained in🧵👇
Nice one
"Gr00t" vs "Pi0" vs "Pi0 Fast". I compared top open-source robotic models, and here's a detailed overview based on our own experience:
"Gr00t" vs "Pi0" vs "Pi0 Fast". I compared top open-source robotic models, and here's a detailed overview based on our own experience:
Robot Learning has focused a lot of the policy. But how about the hardware? Can hardware be learned as well? We will discuss this in the new workshop at RSS.
Excited to announce the 1st Workshop on Robot Hardware-Aware Intelligence @ #RSS2025 in LA! We’re bringing together interdisciplinary researchers exploring how to unify hardware design and intelligent algorithms in robotics! Full info: rss-hardware-intelligence.github.io @RoboticsSciSys
You are Jerry. Who's Tom?
Test-Time Training (TTT) is available in video generation now! We can directly generate complete one-minute video, with great temporal and spatial coherence. We created more episodes of Tom and Jerry (my favorite cartoon in childhood) with our model. test-time-training.github.io/video-dit/…
It's All in the Hips Can we learn anything from the evolution of hip kinematics in humanoid bots? Is there a growing consensus of the optimal design? Yes. And there is data to prove it. First, a primer on Hip Kinematics 🧵
Teleoperation is so tedious. Can we find a better way to scale real-world data? Answer: Human videos. Inspiration: When we were doing teleoperation, we observed the motion performed by the human is almost the same as the robot, it is really just a 3D transformation away. Besides…
Diverse training data leads to a more robust humanoid manipulation policy, but collecting robot demonstrations is slow. Introducing our latest work, Humanoid Policy ~ Human Policy. We advocate human data as a scalable data source for co-training egocentric manipulation policy.⬇️