Jin Cheng
@catachiii
Doctoral Student with @crl_ethz @ETH, working on RL, robotics, and more | Previously @MPI_IS @Tsinghua_Uni - Cells interlinked within cells, interlinked.
🐕 I'm happy to share my paper: RAMBO: RL-augmented Model-based Whole-body Control for Loco-manipulation has been accepted by IEEE Robotics and Automation Letters (RA-L) 🧶 Project website: jin-cheng.me/rambo.github.i… Paper: arxiv.org/abs/2504.06662
🧠With the shift in humanoid control from pure RL to learning from demonstrations, we take a step back to unpack the landscape. 🔗breadli428.github.io/post/lfd/ 🚀Excited to share our blog post on Feature-based vs. GAN-based Learning from Demonstrations—when to use which, and why it…
✨Introducing SENSEI✨ We bring semantically meaningful exploration to model-based RL using VLMs. With intrinsic rewards for novel yet useful behaviors, SENSEI showcases strong exploration in MiniHack, Pokémon Red & Robodesk. Accepted at ICML 2025🎉 Joint work with @cgumbsch 🧵
Like this a lot! It’s actually a good time to explore non-programming way of interacting with robots. Stay tuned we have something out soon.
We interact with dogs through touch -- a simple pat can communicate trust or instruction. Shouldn't interacting with robot dogs be as intuitive? Most commercial robots lack tactile skins. We present UniTac: a method to sense touch using only existing joint sensors! [1/5]
Last Friday, we participated in the annual ETH RobotX Innovation Day. The event began with an inspiring opening by Prof. Stelian Coros. We were proud to share our latest research with the broader ETH robotics community and industry partners.
We are very happy to present our recent work on steerable imitation controllers for legged robots! All Go2s are chilling and happy during the experiments 🐩 Huge congrats and thanks to @eastskykang!
Introducing my recent work, "Learning Steerable Imitation Controllers From Unstructured Animal Motions". In this work, we present a control framework for legged robots that leverages unstructured real-world animal motion data to generate animal-like and user-steerable behaviors.
🎓 Another doctor! Flavio De Vincenti has successfully defended his doctoral thesis on Optimal and Learning-Based Control for Collaborative Loco-Manipulation! Congratulations!
Backflip is nothing new thesedays… but making it work still feels great.
Nice post!
Q-learning is not yet scalable seohong.me/blog/q-learnin… I wrote a blog post about my thoughts on scalable RL algorithms. To be clear, I'm still highly optimistic about off-policy RL and Q-learning! I just think we haven't found the right solution yet (the post discusses why).
Proud to share our latest work from @leggedrobotics ! A legged robot that rallies with humans in badminton 🏸📷 Whole-body agility, onboard perception, RL-trained.
A legged mobile manipulator trained to play badminton with humans coordinates whole-body maneuvers and onboard perception. Paper: science.org/doi/10.1126/sc……Video: youtu.be/zYuxOVQXVt8 @Yuntao144, Andrei Cramariuc, Farbod Farshidian, Marco Hutter
Happy to finally release my last PhD work, where we once again pushed the limits of what ANYmal could do. Now that the PhD at @ETH/ @leggedrobotics is done, we are moving on to robots with fewer legs and more arms at @FlexionRobotics. New robots, same drive to push the limits!
ANYmal just learned parkour in the wild! 9 expert skills → 1 foundation policy → RL fine-tuning → any terrain with a single end-to-end policy from cameras to motor control. Paper: arxiv.org/abs/2505.11164 Watch: youtu.be/QDU_FicBPDo @rdn_nikita Junzhe He Joshua Aurand
EWRL News!⚡ We’re pleased to announce a new Call for Contributed Talks to bring novel and diverse viewpoints to EWRL2025. Early-career researchers are especially encouraged to submit proposals and share their work with the community. Full details: …p-on-reinforcement-learning.github.io/ewrl18/
Slides are here if you are interested: drive.google.com/file/d/1xta-H2…
I am giving a talk "From Sim2Real 1.0 to 4.0 for Humanoid Whole-Body Control and Loco-Manipulation" at the RoboLetics 2.0 workshop @ieee_ras_icra today, summarizing my recent thoughts on sim2real. If you are interested: 2pm, May 23 @ room 302.
📈 Historic Finish! Our lab placed 299th out of 1,106 in SOLA-Stafette 2025! Congratulations to all participants and a big thank you to our wonderful organizer, @NriaArmengol2!