Jianwei Liu

Current PhD Student @rpl_as_ucl and @ai_ucl @FAICDT1;
MEng Biomedical Engineering Alumnus @ImperialBioeng

[Jianwei Liu @JianweiLiu93]

A construction robotics project that I worked on for approx. 2 years has finally been made public.
#robotics #construction #mobilemanipulator #goros #bim #autonomousrobots
https://t.co/7vKUOMP5kY

[Jianwei Liu @JianweiLiu93]

@Em_Nomadic cool figure! is it from a particular paper? kinda reminds me of the mountain goat-inspired robotic feet works done by @thrishlab and @Sariadela_abad 's groups

[Jonathan (Buzz) Embley-Riches @JonEmbleyRiches]

Releasing the alpha of Unreal Robotics Lab — an open-source Unreal Engine plugin with full MuJoCo physics.

Photorealistic rendering and accurate contact physics. No compromises on either side.

GitHub: https://t.co/GAvwsncqtG
Paper: https://t.co/w8sbILR7dW

[Jianwei Liu @JianweiLiu93]

@pravsels @icostafilho 😁

[Nicholas Pfaff @NicholasEPfaff]

Meet SceneSmith: An agentic system that generates entire simulation-ready environments from a single text prompt.

VLM agents collaborate to build scenes with dozens of objects per room, articulated furniture, and full physics properties.

We believe environment generation is no longer the bottleneck for scalable robot training and evaluation in simulation.

Website: https://t.co/UZklSkJe9V

👇🧵(1/8)

[Wenzhe Cai @WenzheC7616]

🤖Can robots achieve accurate navigation without any external localization feedback?
📸We present #LoGoPlanner, which handles perception, localization, and planning in one go!

Check our results on LeKiWi, G1, and Go2 robots.
🌐Project: https://t.co/mWIMzIfuqT

[Sergey Levine @svlevine]

Exciting new work from @ajwagenmaker and collaborators: if we know we will run RL later, we can pretrain with BC (eg for a diffusion VLA) in a way that promotes downstream exploration!

[Matthias Niessner @MattNiessner]

📢📢📢𝐌𝐞𝐬𝐡𝐑𝐢𝐩𝐩𝐥𝐞: Structured Autoregressive Generation of Artist-Meshes

High-fidelity, topologically complete 3D assets that expand naturally like a ripple on a surface! 🌊

Existing AR models often rely on sliding-window inference over truncated segments. However, this limitation breaks long-range geometric dependencies, causing holes and fragmentation.

Instead, MeshRipple uses frontier-aware BFS and sparse-attention global memory to ensure coherent growth with an unbounded receptive field.

-> Highly detailed-mesh generations
-> Artist-like meshing quality
-> Works on room-scale environments

🌍https://t.co/FPmo9QBTac
🎥https://t.co/oV1zBua5iC

Great work by Junkai Lin, Hang Long, Huipeng Guo, Jielei Zhang, Jiayi Yang, Tianle Guo, Yang Yang, Jianwen Li, Wenxiao Zhang, Wei Yang

[Abhishek Gupta @abhishekunique7]

Excited to put out new work - PolaRiS, a framework for scalable generalist policy evaluation!

The idea is simple - short videos of scenes get converted into high-fidelity simulation environments that match the real world. Then you can evaluate your favorite generalist policy on entirely unseen environments purely in simulation, without requiring real-world evaluations 🪇!

Simple right? - turns out getting it to really work needs some careful research and engineering. Let’s investigate! (1/8)

https://t.co/bQdb0aCEY3

[Kyber Labs @KyberLabsRobots]

Demo showing our system doing autonomous assembly of a part! What else should we have it do?

And if you want a hand, wait list is open here!
https://t.co/bh9Xv05eje

[Qiyang (Colin) Li @qiyang_li]

Action chunking is drawing growing interest in RL, yet its theoretical properties are still understudied.

We are excited to share some insights on when we should use action chunking in Q-learning + a new algo (DQC) to tackle hard long-horizon tasks!https://t.co/izVWQBgH3c🧵1/N

[Jianwei Liu @JianweiLiu93]

@AjdDavison Oh cool! Will check these out!

[Jianwei Liu @JianweiLiu93]

@AjdDavison Also here is a slide that I quite like from Chris Eliasmith's group at university of Waterloo, it give a rough energy efficiency scale to compare different compute arch, if they were to be scaled to match the human brain's number of neurons and synapses.

[Jianwei Liu @JianweiLiu93]

@AjdDavison and these Neuromorphic chips are mostly designed for edge compute as well, which are probably more suitable than the mostly server grade Graphcore IPU for robotics applications

[Xiaolong Wang @xiaolonw]

Most robot learning has focused on simple position control. But think about how a human uses a wrench 🔧: you’re not just rotating in one direction—you’re continuously shaping the forces, pushing and pulling differently as you move. Our robot can do exactly that now.

[Ken Goldberg @Ken_Goldberg]

Agreed about occlusions and current datasets being so tiny. I’m concerned about how we’ll get the data needed to solve RLBench problems…I think we need much more research on sim to close the gap for manipulation.

[Winston Gao @WinstonGu_]

Imagine a future where you can ask humanoid robots to clean your room, but some items, like heavy sofas, are too challenging for just one robot to move.

Introducing CooHOI, a learning-based framework designed for the cooperative transportation of objects by multiple humanoid robots. 🤖🤼🤖

Our work has been accepted as Spotlight at NeurIPS 2024. Website: https://t.co/gWFSYEqAAD

[Guanya Shi @GuanyaShi]

Humanoid "wall-flip" motion unlocked. Max linear velocity: 3.5m/s, max angular velocity: 860 degrees per second!

[Jason Peng @xbpeng4]

Implementing motion imitation methods involves lots of nuisances. Not many codebases get all the details right. So, we're excited to release MimicKit!
https://t.co/7enUVUkc3h
A framework with high quality implementations of our methods: DeepMimic, AMP, ASE, ADD, and more to come!