Ireland

Cambridge-Ireland International Centre Research Group

A bridge for predictive physical intelligence: world models, 4D robot vision, generative dynamics, and robot learning across Cambridge and Ireland.

Back to PIRLab groups

Purpose

Prediction as the middle layer between perception and action.

This group develops robot world models that forecast how 3D scenes, objects, and robot states evolve under time and action. The aim is practical: make prediction useful for planning, manipulation, simulation, and safe deployment.

The group is designed as an international collaboration node, connecting physical AI work in Cambridge with research activity in Ireland, including the University of Galway ecosystem.

4D

Temporal robot vision for dynamic, multi-view, and action-conditioned scenes.

3D

Point clouds, neural fields, Gaussian representations, and flow-based dynamics.

Real

World models evaluated through robot tasks, not only offline prediction metrics.

Group Members

Guangming Wang

Group Lead

Dr Guangming Wang

Assistant Professor at the Cambridge-Ireland International Centre and Research Associate at the University of Cambridge; robot world models, physical AI, predictive 3D vision, and engineering automation.

Zhishuo Liu

Cambridge Student

Zhishuo Liu

PhD student at Cambridge.

Nandeesh

Cambridge Student

Nandeesh

PhD student at Cambridge.

Ting Li

Cambridge Student

Ting Li

MPhil student at TUM.

Xinghui Tao

Cambridge Student

Xinghui Tao

PhD student at Cambridge.

Caroline Wendlandt

Cambridge Student

Caroline Wendlandt

MPhil student at TUM.

Sean Deng

Cambridge Student

Sean Deng

Undergraduate student at Cambridge.

Xiaoyu Zhang

Cambridge Student

Xiaoyu Zhang

Undergraduate student at Cambridge.

Jiucai

Cambridge Student

Jiucai

PhD student at Cardiff.

Yucheng

Cambridge Student

Yucheng

PhD student at Tsinghua University.

Yunxiang

Cambridge Student

Yunxiang

PhD student at Tsinghua University.

Yixiong Jing

Group Co-Lead

Dr Yixiong Jing

Postdoctoral Research Associate at the University of Cambridge; 3D reconstruction, generative models, physical reasoning with 3DGS, and robotics.

Haibing Wu

Group Co-Lead

Haibing Wu

PhD Candidate in the CV4DT Research Group.

Wanru Yang

Cambridge Student

Wanru Yang

PhD student at Cambridge.

Chenhong Zhang

Researcher

Chenhong Zhang

PhD Candidate in the CV4DT Research Group.

Advisory Board

Cambridge-Ireland advisory board.

Laing O'Rourke Associate Professor in Construction Engineering at Cambridge; intelligent underground construction and monitoring.

Full Professor of Computer Science at ETH Zurich and Director of the Microsoft Mixed Reality and AI Lab in Zurich; 3D computer vision, robot vision, SLAM, and semantic scene understanding.

Cheryl and John Neerhout, Jr. Distinguished Professor and Distinguished Professor of Mechanical Engineering at UC Berkeley; adaptive control, digital control, mechatronics, and robotics.

Representative work

Predictive models for embodied intelligence.

Diffusion scene flow pipeline

Prediction

DifFlow3D: Toward Robust Uncertainty-Aware Scene Flow Estimation with Diffusion Model

Diffusion-based refinement for robust 3D scene flow under dynamic, noisy, and ambiguous cases.

3D scene flow pseudo-labelling visualization

Prediction

3DSFLabelling: Boosting 3D Scene Flow Estimation by Pseudo Auto-labelling

Auto-labelling methods for richer motion supervision in real 3D scenes.

Real2Sim2Real manipulation learning

Prediction to action

RL-GSBridge: 3D Gaussian Splatting Based Real2Sim2Real Method for Robotic Manipulation Learning

A real-to-sim-to-real pipeline that uses 3D Gaussian Splatting inside robot reinforcement learning.

Join

Build robot world models with us.

Relevant backgrounds include embodied AI, video prediction, 3D generative models, robot learning, reinforcement learning, simulation, and real robot experimentation.

Contact