Zhen Fan
M.Sc. in Mechanical Engineering, National University of Singapore (NUS)
Research Interests: Human–Robot Collaboration, Vision–Language–Action Models, Simulation-to-Real Transfer, Explainable Robotics
Research Interests: Human–Robot Collaboration, Vision–Language–Action Models, Simulation-to-Real Transfer, Explainable Robotics
About
I am interested in building human-like collaboration systems between humans and robots. My research focuses on integrating multimodal perception, high-level reasoning, and structured semantic representations to enable interpretable and reliable robot decision-making in real-world manufacturing scenarios.
Recently, I have worked on Omniverse-based digital twins and automated synthetic-data pipelines for metal additive manufacturing, multimodal knowledge graphs for technical documentation, and vision–language–action models for human–robot collaboration.
Research Highlights
- Human–Robot Collaboration (HRC): XR/digital-twin interfaces that allow humans to teach, supervise and recover robot behaviors.
- Vision–Language–Action (VLA): leveraging large multimodal models to ground task semantics and robot skills in real factories.
- Sim2Real and Synthetic Data: using high-fidelity simulators and synthetic videos to reduce data collection cost while preserving real-world performance.
- Explainable Robotics: task semantic graphs and knowledge graphs as a safety and explanation layer on top of robot policies.
Selected Projects
- AMSim2Real – Simulation-driven human action recognition for metal additive manufacturing, bridging virtual demonstrations and real operators on a metal 3D printer.
- MetalMind – Knowledge-graph-enhanced digital twins that connect technical documents, human procedures and robot skills in industrial environments.
See the Projects page for more details.
News
- [2025] Two journal papers published on human-centric knowledge systems and digital twins for metal additive manufacturing.
- [2025] Building a synthetic video dataset for VLM/VLA training in manufacturing scenarios.