Daniele Panozzo
New York University, NY·Computer Science
Engineers and scientists often build a digital twin, a computer model of a real object, so they can study it without costly or risky physical experiments. Making these models useful raises two hard problems.
- Fields
Computer Science
Engineering
Mathematics & Statistics
- Topics
differentiable simulation
digital twins
finite element method
- Grants
- 1 known grant, through Aug 2029
What they work on
- Develop dPolyFEM, an open-source modular framework for differentiable simulation, which computes how simulation results change with inputs, for elastic objects with large deformations, contact and friction.
- Treat meshing, the finite-element solution and the adjoint sensitivity calculation as a single end-to-end differentiable problem, so the software handles imperfect meshes and shapes without manual parameter tuning.
- Recast the time-stepping of physical systems as energy minimization problems, and build robust, parallel solvers that let users balance speed against accuracy.
- Evaluate the software on large-scale benchmarks and on applications in computational fabrication, patient-specific biomechanics and robotics, including inferring material parameters and optimizing shapes.
More about the research
The first is tuning the model to match real-world measurements. The second is changing a design until it meets a goal. Differentiable simulation helps with both. It computes how a simulation's output changes when its inputs change, so the computer can adjust the inputs automatically. Two examples are finding the realistic material properties of a patient-specific body model and finding the shoe-sole shape that spreads load evenly.
Why it matters
Free software that automatically fits simulations to real data and optimizes designs could reduce the need for expensive physical prototypes and experiments. Engineers could find better shapes for manufactured parts such as shoe soles. Biomechanics researchers could build patient-specific body models with realistic material properties. Robotics developers could improve control of soft, deforming systems. The modules will be integrated into major open-source tools that potentially serve tens of thousands of users.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Mallikarjunaiah Muddamallappa
Texas A & M University-Corpus Christi, TX·Mathematics & Statistics
Formulate a regularized phase-field model of brittle fracture that uses strain-limiting constitutive relations, avoiding the unrealistic infinite strain predicted by linear models at crack tips.
David Andrew Hyde
Vanderbilt University, TN·Computer Science·Early-career
Develop optimized GPU implementations of grids, particles, octrees and linear solvers as the foundation of COSTA, an open cloud platform for physics simulation.
Lin Mu
University of Georgia, GA·Mathematics & Statistics
Design a finite element scheme for incompressible flow that achieves pressure robustness and exact mass conservation at once, using a divergence-preserving discretization rather than the trade-off made by traditional solvers.
The school
New York University: 72 labs on PI Pages, $1.5B a year on research, 540 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.