×2
×2
×3
×1
×1
×1
×1
×2
×2
×3
×1
×1
×1
×1
×2
×2
×1
×1
×1
×1
×3
×2
×2
×1
×1
×1
×1
×3
×2 Welcome to the Multimodal embodied AI and Robotic Systems (MARS) Lab at Nanyang Technological University (NTU), Singapore. We seek to build reliable embodied intelligence in the physical world.
Our research is driven by a central question: How can robots learn and interact with the world as robustly and adaptively as humans? To address this challenge, we develop multimodal embodied AI systems and feedback learning to understand physical interactions and acquire complex skills. We study the full stack of robotics, from multimodal perception models to world models, robot learning, decision making, and real-world deployment.
Our long-term goal is to build trustworthy physical AI systems that can continually learn from experience, collaborate naturally with humans, and operate reliably in homes, healthcare, manufacturing, and scientific laboratories.
News
- Sep 2026 Four papers are accepted by CoRL 2026. Congratulations to Bofan Lyu, Xinying Guo, Chenxi Jiang, Xinyu Zhou and Zhe Li!
- Aug 2026 Our paper Physics Filtering is accepted by Nature Portfolio Robotics. Congratulations to Jindou Jia!
- Aug 2026 Our survey World Model for Robot Learning is accepted by IJRR. Congratulations to Bohan Hou, Gen Li, Jindou Jia, Tuo An and Xinying Guo!
- Apr 2026 Our paper A2A is accepted by RSS 2026. Congratulations to Jindou Jia and Gen Li!
- Mar 2026 We received an NVIDIA Academic Grant on Robotics.
- Feb 2026 Three papers (two main + one findings) are accepted by CVPR 2026.
- Feb 2026 We received a Microsoft Seed Grant from Microsoft Research Asia.
- Jan 2026 Our real-world reinforcement learning paper is accepted by ICRA 2026. Congratulations to Xiangyu Chen!
- Jan 2026 Two papers are accepted by ICLR 2026. Congratulations to Xinyan Chen and Chenxi Jiang!
Events
Collaborators