A virtual environment for **differentiable elastoplastic** soft object manipulation.
A data-oriented programming language and compiler for compute-intense visual computing workloads, especially those needing **spatially sparse data structures**.
Deep learning control and state representation simultaneously for soft robots.
The first differentiable MLS-MPM simulator.
MIT-thesis version of [ChainQueen](publication/2019-chainqueen)
High-performance giga-voxel FEM solver for topology optimization on a narrow-band, sparse grid data structure.
Use deep learning to generate faithful spectral color reproduction for 3D printing.
Differentiable photo-editing model with RL and GANs.
The moving least squares material point method (MLS-MPM) that runs fast fast and is easy to implement.
The asynchronous material point method (AsyncMPM): Use small time steps where necessary, and large time steps where possible.