Multimodal AI and Robotic Systems (MARS) Lab is led by Prof. Jianfei Yang in Nanyang Technological University, Singapore.
MARS Lab studies Physical AI, focusing on how AI can enable physical systems—such as robotics, IoT, and industrial systems—to perceive, understand, and interact with the physical world.
News
- May 2025: (member) Our first humanoid robot member “MARS-1” is coming to the lab.
- Feb 2025: (project) Our lab established a collaboration with Abio for robotic lab automation.
- Feb 2025: (funding) Our project is supported by Microsoft Accelerating Foundation Models Research.
- Feb 2025: (paper) Our paper (Feedback Neural Network) is selected as an ICLR oral paper (top 1.8%).
- Jan 2025: (paper) Two papers (on X-Fi and Feedback Neural Network) are accepted by ICLR-25.
- Jan 2025: (paper) One paper on point registration is accepted by ICRA-25. Congrats to our undergrad intern, Geng Li!
- Jan 2025: (project) Our lab participates in Project Aria by Meta!
- Sep 2024: (award) The 1st PhD student, Zhou Chuhao, won the first place at ECCV-24 Data Distillation Challenge. Congratulations!
- Sep 2024: (kick-off) Prof. Jianfei Yang launched MARS Lab at the Robotics Research Center, NTU.
Highlights

Physical AI Research
At MARS Lab, we mainly study Human-Centric Physical AI and Robotics, ranging from multimodal perception, human-robot interaction, robotic manipulation, and embodied locomotion with AIoT integration, enabling intelligent machines to perceive, act, and adapt seamlessly in the physical world.

Research Achievements
Our MARS Lab embodies the open-source spirit, accelerating scientific progress through impactful libraries such as SenseFi and MM-Fi, which support multimodal foundation models. We develop AI algorithms and deploy them in real-world robotic and AIoT systems. Our research is widely recognized, with publications in top venues across AI (NeurIPS, ICLR, CVPR, ICCV, ACL, TPAMI, IJCV), robotics (ICRA, IROS), and interdisciplinary science (Cell Patterns, Briefings in Bioinformatics, CACIE).

Our Team
MARS Lab is a diverse and close-knit team of researchers from various academic and cultural backgrounds. Equipped with advanced robotic platforms, including multiple robotic arms and the Unitree G1 humanoid. We work together in a collaborative, supportive environment to push the boundaries of AI and robotics through shared curiosity and innovation.