CLEAR Lab

The Control & Learning for Robotics and Autonomy (CLEAR) Lab, led by Prof. Wei Zhang, focuses on advancing robotics and autonomous systems.

[LimX Dynamics @LimX_Dynamics]

See LimX COSA 0.5 live.
No teleoperation. No cuts. One take.
One of the world's few complete demonstrations of long-horizon mobile manipulation, bringing fully autonomous humanoid robots another step closer to us.
📄Tech report now live on our website.

Join us at WAIC 2026:
Jul 17–20
H3 Hall, Booth C-216, Shanghai World Expo Exhibition & Convention Center
Experience our brain-cerebellum embodied AI stack, including COSA.

[LimX Dynamics @LimX_Dynamics]

LimX COSA 0.5: LimX Dynamics' Humanoid Brain System Updated

No teleoperation. No cuts. Long take. One of the world's few complete demonstrations of long-horizon mobile manipulation, bringing fully autonomous humanoid robots another step closer to us.
#LimXDynamcis #BrainSystem #PhysicalAI #LimXOli #humanoidrobot

[CLEAR Lab @CLEARLabChina]

Multi-Embodiment Robotic Retargeting via Guided Diffusion Model is accepted by CASE 2026.

Arxiv: https://t.co/20w4gA0Ni1

[CLEAR Lab @CLEARLabChina]

Multi-Embodiment Robotic Retargeting via Guided Diffusion Model is accepted by CASE 2026.

Arxiv: https://t.co/20w4gA0Ni1

[LimX Dynamics @LimX_Dynamics]

Awesome having @TheHumanoidHub visit our headquarter at LimX Dynamics in Shenzhen.

From live demos of Oli and TRON to chatting with our Head of R&D, Paul Li, we discussed our approach to humanoids and the fundamental engineering challenges of bringing them into the real world.

It was a great conversation on where humanoids are today and what's next. There's plenty to be excited about, while more to learn through iteration and real-world deployment.

Thanks for the visit. Enjoy the full tour below 👇
#humanoid #EmbodiedAI #robotics #LimXDynamics

[CLEAR Lab @CLEARLabChina]

“PolicyFlow: Policy Optimization with Continuous Normalizing Flow in Reinforcement Learning” is accepted by ICLR 2026.

[CLEAR Lab @CLEARLabChina]

“CWI: Composite Humanoid Whole-Body Imitation System for Loco-manipulation” is accpeted by RAL 2026.

[CLEAR Lab @CLEARLabChina]

“SSI-Policy: Learning Structured Scene Interfaces for Vision-Language Robotic Manipulation” is accepted by IROS 2026

[CLEAR Lab @CLEARLabChina]

📄 Paper: AISPO: Enhancing Depth Reliability for Robotic Manipulation of Non-Lambertian Objects via Affine-Invariant Shape Prior
arXiv: https://t.co/6PySX7Jsz8

[CLEAR Lab @CLEARLabChina]

🤖 Transparent cups, glass bowls, reflective metals…
These objects are everywhere in our daily lives, but they are far from “robot-friendly.” They are not completely invisible to robots, but they often make depth cameras unreliable:

[CLEAR Lab @CLEARLabChina]

while still reconstructing reasonable geometric structures. 👀 When robots see more accurately, they can grasp more stably.

[CLEAR Lab @CLEARLabChina]

Please check out our GitHub page (https://t.co/C454A4qNaj)! More open-source resources are coming soon!