I am a simulation researcher and engineer working at the intersection of physics-based simulation, nonlinear optimization and robot learning. I currently lead the simulation team at a robotics startup in Shenzhen, China, where we build a GPU-accelerated rigid–soft coupled simulation engine and use it to generate large-scale robot interaction data (soft bodies, cloth and articulated objects) for world-model training and dexterous-hand manipulation.

I received my Ph.D. in Computer Graphics from Nanyang Technological University (NTU), Singapore (2019–2025, advised by Prof. Jianmin Zheng), where I worked on physics-based simulation and material-distribution design for 3D printing. Before that I obtained an M.Eng. in Electrical Engineering from the National University of Singapore (NUS) and a B.Eng. in Communication Engineering from the University of Electronic Science and Technology of China (UESTC).

Research interests

• Physics-based and differentiable simulation: FEM based on continuum mechanics, PBD/XPBD, IPC, MPM and projective dynamics; contact and large-deformation handling; GPU-parallel solvers (C++/CUDA).
• Simulation for robot learning: rigid–soft coupled simulation engines integrated with Isaac Lab, synthetic data generation for world models and reinforcement learning, perception–world-model–control integration for fine manipulation in everyday scenes.
• Nonlinear optimization and AI for design: large-scale discrete material-distribution optimization (L0 regularization, graph cuts, ADMM), topology optimization with CNNs and physics-informed neural networks, generative microstructure design.

Current work

At the company I lead a seven-person team responsible for rigid-body, deformable-body and cloth simulation. Highlights so far:

• Designed and built a fully GPU-based rigid–soft coupled simulation architecture with a USD parsing layer, and migrated the simulation pipeline to the USD framework.
• Generated high-quality synthetic data for rigid–soft coupled scenes such as parcel grasping and hanging garments, and validated advanced grasping algorithms on elasto-plastic objects.
• Integrated the engine into Isaac Lab for penetration-free simulation of robot arms grasping cloth and elastic bodies, with Manager-Based Env support for PPO training.
• Since July 2026: large-scale generation of robot interaction data for world-model training, and an integrated perception + world model + action-control stack for fine-grained dexterous-hand control in daily-life scenarios.

Selected publications

• H. Li, W. Zhang, J. Zheng, E. D. Davis, J. Zeng. Optimizing heterogeneous elastic material distributions on 3D models. Computer-Aided Design, 175, 103748, 2024. DOI
• H. Li, J. Zheng. L0-regularization based material design for hexahedral mesh models. Computer-Aided Design and Applications, 19(6), 1171–1183, 2022. DOI
• J. Zheng, H. Li. Designing multi-material distributions in 3D parts for desired deformation behaviour. Dagstuhl Seminar 24241, Dagstuhl Reports, 14(6), 2024. DOI

See the Publications and CV pages for the full list.