arXiv subjects

Eitan Grinspun

Publications and source records attributed to Eitan Grinspun.

At least 19 recordsLinked to original sources

arXiv2026

Factorized Neural Implicit DMD for Parametric Dynamics

A data-driven, model-free approach to modeling the temporal evolution of physical systems mitigates the need for explicit knowledge of the governing equations. Even when physical priors such as partial differential equations are available, such systems often reside in high-dimensional state spaces and exhibit nonlinear dynamics, making traditional numerical solvers computationally expensive and ill-suited for real-time analysis and control. Consider the problem of learning a parametric flow of a dynamical system: with an initial field and a set of physical parameters, we aim to predict the system's evolution over time in a way that supports long-horizon rollouts, generalization to unseen parameters, and spectral analysis. We propose a physics-coded neural field parameterization of the Koopman operator's spectral decomposition. Unlike a physics-constrained neural field, which fits a single solution surface, and neural operators, which directly approximate the solution operator at fixed time horizons, our model learns a factorized flow operator that decouples spatial modes and temporal evolution. This structure exposes underlying eigenvalues, modes, and stability of the underlying physical process to enable stable long-term rollouts, interpolation across parameter spaces, and spectral analysis. We demonstrate the efficacy of our method on a range of dynamics problems, showcasing its ability to accurately predict complex spatiotemporal phenomena while providing insights into the system's dynamic behavior.

cs.LG↗

arXiv2026

The Configurational Element Method for Nonconvex Granular Media

Granular media surround us, comprising everything from the ground we walk on to the foods we eat. Owing to their ubiquity our ability to understand and predict the mechanical evolution of grains is not only of key scientific importance, but is also a key component to synthesize believable animations of our world. Despite their importance, shortcomings persist in our ability to simulate granular media. In particular, simulating grains with non-convex shapes remains a challenging and computationally expensive task. We propose a method to simulate non-convex rigid grains by posing geometric contact in configuration space and learning the resulting contact map with a neural network. Our formulation reduces the complex task of modeling and simulating non-convex shapes to simple network evaluations that are easily run on standard compute hardware, allowing us to quickly and robustly simulate large scale systems of non-convex grains.

cs.GR↗

arXiv2026

Variational Green's Functions for Volumetric PDEs

Green's functions characterize the fundamental solutions of partial differential equations; they are essential for tasks ranging from shape analysis to physical simulation, yet they remain computationally prohibitive to evaluate on arbitrary geometric discretizations. We present Variational Green's Function (VGF), a method that learns a smooth, differentiable representation of the Green's function for linear self-adjoint PDE operators, including the Poisson, the screened Poisson, and the biharmonic equations. To resolve the sharp singularities characteristic of the Green's functions, our method decomposes the Green's function into an analytic free-space component, and a learned corrector component. Our method leverages a variational foundation to impose Neumann boundary conditions naturally, and imposes Dirichlet boundary conditions via a projective layer on the output of the neural field. The resulting Green's functions are fast to evaluate, differentiable with respect to source application, and can be conditioned on other signals parameterizing our geometry.

cs.GR↗

arXiv2026

Low-Rank Koopman Deformables with Log-Linear Time Integration

We present a low-rank Koopman operator formulation for accelerating deformable subspace simulation. Using a Dynamic Mode Decomposition (DMD) parameterization of the Koopman operator, our method learns the temporal evolution of deformable dynamics and predicts future states through efficient matrix evaluations instead of sequential time integration. This yields log-linear scaling in the number of time steps and allows large portions of the trajectory to be skipped while retaining accuracy. The resulting temporal efficiency is especially advantageous for optimization tasks such as control and initial-state estimation, where the objective often depends largely on the final configuration. To broaden the scope of Koopman-based reduced-order models in graphics, we introduce a discretization-agnostic extension that learns shared dynamic behavior across multiple shapes and mesh resolutions. Prior DMD-based approaches have been restricted to a single shape and discretization, which limits their usefulness for tasks involving geometry variation. Our formulation generalizes across both shape and discretization, which enables fast shape optimization that was previously impractical for DMD models. This expanded capability highlights the potential of Koopman operator learning as a practical tool for efficient deformable simulation and design.

cs.GR↗

arXiv2026

Topology- and Geometry-Exact Coupling for Incompressible Fluids and Thin Deformables

We introduce a topology-preserving discretization for coupling incompressible fluids with thin deformable structures, achieving guaranteed leakproofness through preservation of fluid domain connectivity. Our approach leverages a stitching algorithm applied to a clipped Voronoi diagram generated from Lagrangian fluid particles, in order to maintain path connectivity around obstacles. This geometric discretization naturally conforms to arbitrarily thin structures, enabling boundary conditions to be enforced exactly at fluid-solid interfaces. By discretizing the pressure projection equations on this conforming mesh, we can enforce velocity boundary conditions at the interface for the fluid while applying pressure forces directly on the solid boundary, enabling sharp two-way coupling between phases. The resulting method prevents fluid leakage through solids while permitting flow wherever a continuous path exists through the fluid domain. We demonstrate the effectiveness of our approach on diverse scenarios including flows around thin membranes, complex geometries with narrow passages, and deformable structures immersed in liquid, showcasing robust two-way coupling without artificial sealing or leakage artifacts.

physics.comp-ph↗

arXiv2026

Odd-DC: Generalizable Neural Model Reduction via Odd Difference-of-Convex Structure

Model reduction is essential for real-time simulation of deformable objects. Linear techniques such as PCA provide structured and predictable behavior, but their limited expressiveness restricts accuracy under large or nonlinear deformations. Nonlinear model reduction with neural networks offers richer representations and higher compression; however, without structural constraints, the learned mapping from latent coordinates to displacements often generalizes poorly beyond the training distribution. We present an odd difference-of-convex (DC) neural formulation that bridges linear and nonlinear model reduction. Our goal is to obtain a latent space that behaves reliably under unseen load magnitudes and directions. To improve extrapolation in magnitude, we introduce convexity into the decoder to discourage oscillatory responses. Yet convexity alone cannot represent the odd symmetry required by many symmetric systems, which is crucial for generalization to inverse force directions. We therefore adopt a DC formulation that preserves the stabilizing effect of convexity while explicitly enforcing odd symmetry. Practically, we realize this structure using an input-convex neural network (ICNN) augmented with symmetry constraints. Across challenging deformation scenarios with varying magnitudes and reversed load directions, our method demonstrates stronger generalization than unconstrained nonlinear reductions while maintaining compact latent spaces and real-time performance. Our DC formulation extends to both mesh-based and neural-field reductions, demonstrating applicability across multiple classes of neural nonlinear model reduction.

cs.GR↗

arXiv2025

MorphModes: Non-rigid Registration via Adaptive Skinning Eigenmodes

Non-rigid registration is a crucial task with applications in medical imaging, industrial robotics, computer vision, and entertainment. Standard approaches accomplish this task using variations on the Non-Rigid Iterative Closest Point (NRICP) algorithms, which are prone to local minima and sensitive to initial conditions. We instead formulate the non-rigid registration problem as a Signed Distance Function (SDF) matching optimization problem, which provides richer shape information compared to traditional ICP methods. To avoid degenerate solutions, we propose to use a smooth Skinning Eigenmode subspace to parameterize the optimization problem. Finally, we propose an adaptive subspace optimization scheme to allow the resolution of localized deformations within the optimization. The result is a non-rigid registration algorithm that is more robust than NRICP, without the parameter sensitivity present in other SDF-matching approaches.

cs.GR↗

arXiv2025

Fast Subspace Fluid Simulation with a Temporally-Aware Basis

We present a novel reduced-order fluid simulation technique leveraging Dynamic Mode Decomposition (DMD) to achieve fast, memory-efficient, and user-controllable subspace simulation. We demonstrate that our approach combines the strengths of both spatial reduced order models (ROMs) as well as spectral decompositions. By optimizing for the operator that evolves a system state from one timestep to the next, rather than the system state itself, we gain both the compressive power of spatial ROMs as well as the intuitive physical dynamics of spectral methods. The latter property is of particular interest in graphics applications, where user control of fluid phenomena is of high demand. We demonstrate this in various applications including spatial and temporal modulation tools and fluid upscaling with added turbulence. We adapt DMD for graphics applications by reducing computational overhead, incorporating user-defined force inputs, and optimizing memory usage with randomized SVD. The integration of OptDMD and DMD with Control (DMDc) facilitates noise-robust reconstruction and real-time user interaction. We demonstrate the technique's robustness across diverse simulation scenarios, including artistic editing, time-reversal, and super-resolution. Through experimental validation on challenging scenarios, such as colliding vortex rings and boundary-interacting plumes, our method also exhibits superior performance and fidelity with significantly fewer basis functions compared to existing spatial ROMs. The inherent linearity of the DMD operator enables unique application modes, such as time-reversible fluid simulation. This work establishes another avenue for developing real-time, high-quality fluid simulations, enriching the space of fluid simulation techniques in interactive graphics and animation.

cs.GR↗

arXiv2025

Script2Screen: Supporting Dialogue Scriptwriting with Interactive Audiovisual Generation

Scriptwriting has traditionally been text-centric, a modality that only partially conveys the produced audiovisual experience. A formative study with professional writers informed us that connecting textual and audiovisual modalities can aid ideation and iteration, especially for writing dialogues. In this work, we present Script2Screen, an AI-assisted tool that integrates scriptwriting with audiovisual scene creation in a unified, synchronized workflow. Focusing on dialogues in scripts, Script2Screen generates expressive scenes with emotional speeches and animated characters through a novel text-to-audiovisual-scene pipeline. The user interface provides fine-grained controls, allowing writers to fine-tune audiovisual elements such as character gestures, speech emotions, and camera angles. A user study with both novice and professional writers from various domains demonstrated that Script2Screen's interactive audiovisual generation enhances the scriptwriting process, facilitating iterative refinement while complementing, rather than replacing, their creative efforts.

cs.HC↗

arXiv2025

Lifting the Winding Number: Precise Discontinuities in Neural Fields for Physics Simulation

Cutting thin-walled deformable structures is common in daily life, but poses significant challenges for simulation due to the introduced spatial discontinuities. Traditional methods rely on mesh-based domain representations, which require frequent remeshing and refinement to accurately capture evolving discontinuities. These challenges are further compounded in reduced-space simulations, where the basis functions are inherently geometry- and mesh-dependent, making it difficult or even impossible for the basis to represent the diverse family of discontinuities introduced by cuts. Recent advances in representing basis functions with neural fields offer a promising alternative, leveraging their discretization-agnostic nature to represent deformations across varying geometries. However, the inherent continuity of neural fields is an obstruction to generalization, particularly if discontinuities are encoded in neural network weights. We present Wind Lifter, a novel neural representation designed to accurately model complex cuts in thin-walled deformable structures. Our approach constructs neural fields that reproduce discontinuities precisely at specified locations, without baking in the position of the cut line. Crucially, our approach does not embed the discontinuity in the neural network's weights, opening avenues to generalization of cut placement. Our method achieves real-time simulation speeds and supports dynamic updates to cut line geometry during the simulation. Moreover, the explicit representation of discontinuities makes our neural field intuitive to control and edit, offering a significant advantage over traditional neural fields, where discontinuities are embedded within the network's weights, and enabling new applications that rely on general cut placement.

cs.GR↗

arXiv2025

Force-Dual Modes: Subspace Design from Stochastic Forces

Designing subspaces for Reduced Order Modeling (ROM) is crucial for accelerating finite element simulations in graphics and engineering. Unfortunately, it's not always clear which subspace is optimal for arbitrary dynamic simulation. We propose to construct simulation subspaces from force distributions, allowing us to tailor such subspaces to common scene interactions involving constraint penalties, handles-based control, contact and musculoskeletal actuation. To achieve this we adopt a statistical perspective on Reduced Order Modelling, which allows us to push such user-designed force distributions through a linearized simulation to obtain a dual distribution on displacements. To construct our subspace, we then fit a low-rank Gaussian model to this displacement distribution, which we show generalizes Linear Modal Analysis subspaces for uncorrelated unit variance force distributions, as well as Green's Function subspaces for low rank force distributions. We show our framework allows for the construction of subspaces that are optimal both with respect to physical material properties, as well as arbitrary force distributions as observed in handle-based, contact, and musculoskeletal scene interactions.

cs.GR↗

arXiv2025

Topology-Preserving Coupling of Compressible Fluids and Thin Deformables

We present a novel discretization of coupled compressible fluid and thin deformable structures that provides sufficient and necessary leakproofness by preserving the path connectedness of the fluid domain. Our method employs a constrained Voronoi-based spatial partitioning combined with Godunov-style finite-volume time integration. The fluid domain is discretized into cells that conform exactly to the fluid-solid interface, allowing boundary conditions to be sharply resolved exactly at the interface. This enables direct force exchange between the fluid and solid while ensuring that no fluid leaks through the solid, even when arbitrarily thin. We validate our approach on a series of challenging scenarios -- including a balloon propelled by internal compressed air, a champagne cork ejecting after overcoming friction, and a supersonic asteroid -- demonstrating bidirectional energy transfer between fluid and solid.

physics.comp-ph↗

arXiv2025

Precise Gradient Discontinuities in Neural Fields for Subspace Physics

Discontinuities in spatial derivatives appear in a wide range of physical systems, from creased thin sheets to materials with sharp stiffness transitions. Accurately modeling these features is essential for simulation but remains challenging for traditional mesh-based methods, which require discontinuity-aligned remeshing -- entangling geometry with simulation and hindering generalization across shape families. Neural fields offer an appealing alternative by encoding basis functions as smooth, continuous functions over space, enabling simulation across varying shapes. However, their smoothness makes them poorly suited for representing gradient discontinuities. Prior work addresses discontinuities in function values, but capturing sharp changes in spatial derivatives while maintaining function continuity has received little attention. We introduce a neural field construction that captures gradient discontinuities without baking their location into the network weights. By augmenting input coordinates with a smoothly clamped distance function in a lifting framework, we enable encoding of gradient jumps at evolving interfaces. This design supports discretization-agnostic simulation of parametrized shape families with heterogeneous materials and evolving creases, enabling new reduced-order capabilities such as shape morphing, interactive crease editing, and simulation of soft-rigid hybrid structures. We further demonstrate that our method can be combined with previous lifting techniques to jointly capture both gradient and value discontinuities, supporting simultaneous cuts and creases within a unified model.

cs.GR↗

arXiv2025

Shape Space Spectra

Eigenanalysis of differential operators, such as the Laplace operator or elastic energy Hessian, is typically restricted to a single shape and its discretization, limiting reduced order modeling (ROM). We introduce the first eigenanalysis method for continuously parameterized shape families. Given a parametric shape, our method constructs spatial neural fields that represent eigenfunctions across the entire shape space. It is agnostic to the specific shape representation, requiring only an inside/outside indicator function that depends on shape parameters. Eigenfunctions are computed by minimizing a variational principle over nested spaces with orthogonality constraints. Since eigenvalues may swap dominance at points of multiplicity, we jointly train multiple eigenfunctions while dynamically reordering them based on their eigenvalues at each step. Through causal gradient filtering, this reordering is reflected in backpropagation. Our method enables applications to operate over shape space, providing a single ROM that encapsulates vibration modes for all shapes, including previously unseen ones. Since our eigenanalysis is differentiable with respect to shape parameters, it facilitates eigenfunction-aware shape optimization. We evaluate our approach on shape optimization for sound synthesis and locomotion, as well as reduced-order modeling for elastodynamic simulation.

cs.GR↗

arXiv2024

Actuators À La Mode: Modal Actuations for Soft Body Locomotion

Traditional character animation specializes in characters with a rigidly articulated skeleton and a bipedal/quadripedal morphology. This assumption simplifies many aspects for designing physically based animations, like locomotion, but comes with the price of excluding characters of arbitrary deformable geometries. To remedy this, our framework makes use of a spatio-temporal actuation subspace built off of the natural vibration modes of the character geometry. The resulting actuation is coupled to a reduced fast soft body simulation, allowing us to formulate a locomotion optimization problem that is tractable for a wide variety of high resolution deformable characters.

cs.GR↗

arXiv2024

Subspace Mixed-FEM for Real-Time Heterogeneous Elastodynamics

We propose a reduced space mixed finite element method (MFEM) built on a Skinning Eigenmode subspace and material-aware cubature scheme. Our solver is well-suited for simulating scenes with large material and geometric heterogeneities in real-time. This mammoth geometry is composed of 98,175 vertices and 531,565 tetrahedral elements and with a heterogenous composition of widely varying materials of muscles ($E= 5\times10^5$ Pa), joints ($E=1\times10^5$ Pa), and bone ($E=1\times10^{10}$ Pa). The resulting simulation runs at 120 frames per second (FPS).

cs.GR↗

arXiv2023

Neural Stress Fields for Reduced-order Elastoplasticity and Fracture

We propose a hybrid neural network and physics framework for reduced-order modeling of elastoplasticity and fracture. State-of-the-art scientific computing models like the Material Point Method (MPM) faithfully simulate large-deformation elastoplasticity and fracture mechanics. However, their long runtime and large memory consumption render them unsuitable for applications constrained by computation time and memory usage, e.g., virtual reality. To overcome these barriers, we propose a reduced-order framework. Our key innovation is training a low-dimensional manifold for the Kirchhoff stress field via an implicit neural representation. This low-dimensional neural stress field (NSF) enables efficient evaluations of stress values and, correspondingly, internal forces at arbitrary spatial locations. In addition, we also train neural deformation and affine fields to build low-dimensional manifolds for the deformation and affine momentum fields. These neural stress, deformation, and affine fields share the same low-dimensional latent space, which uniquely embeds the high-dimensional simulation state. After training, we run new simulations by evolving in this single latent space, which drastically reduces the computation time and memory consumption. Our general continuum-mechanics-based reduced-order framework is applicable to any phenomena governed by the elastodynamics equation. To showcase the versatility of our framework, we simulate a wide range of material behaviors, including elastica, sand, metal, non-Newtonian fluids, fracture, contact, and collision. We demonstrate dimension reduction by up to 100,000X and time savings by up to 10X.

cs.GR↗

arXiv2023

LiCROM: Linear-Subspace Continuous Reduced Order Modeling with Neural Fields

Linear reduced-order modeling (ROM) simplifies complex simulations by approximating the behavior of a system using a simplified kinematic representation. Typically, ROM is trained on input simulations created with a specific spatial discretization, and then serves to accelerate simulations with the same discretization. This discretization-dependence is restrictive. Becoming independent of a specific discretization would provide flexibility to mix and match mesh resolutions, connectivity, and type (tetrahedral, hexahedral) in training data; to accelerate simulations with novel discretizations unseen during training; and to accelerate adaptive simulations that temporally or parametrically change the discretization. We present a flexible, discretization-independent approach to reduced-order modeling. Like traditional ROM, we represent the configuration as a linear combination of displacement fields. Unlike traditional ROM, our displacement fields are continuous maps from every point on the reference domain to a corresponding displacement vector; these maps are represented as implicit neural fields. With linear continuous ROM (LiCROM), our training set can include multiple geometries undergoing multiple loading conditions, independent of their discretization. This opens the door to novel applications of reduced order modeling. We can now accelerate simulations that modify the geometry at runtime, for instance via cutting, hole punching, and even swapping the entire mesh. We can also accelerate simulations of geometries unseen during training. We demonstrate one-shot generalization, training on a single geometry and subsequently simulating various unseen geometries.

cs.GR↗