We introduce SimAvatar, a framework designed to generate simulation-ready clothed 3D human avatars from a text prompt. Current text-driven human avatar generation methods either model hair, clothing, and the human body using a unified geometry or produce hair and garments that are not easily adaptable for simulation within existing simulation pipelines. The primary challenge lies in representing the hair and garment geometry in a way that allows leveraging established prior knowledge from foundational image diffusion models (e.g., Stable Diffusion) while being simulation-ready using either physics or neural simulators. To address this task, we propose a two-stage framework that combines the flexibility of 3D Gaussians with simulation-ready hair strands and garment meshes. Specifically, we first employ three text-conditioned 3D generative models to generate garment mesh, body shape and hair strands from the given text prompt. To leverage prior knowledge from foundational diffusion models, we attach 3D Gaussians to the body mesh, garment mesh, as well as hair strands and learn the avatar appearance through optimization. To drive the avatar given a pose sequence, we first apply physics simulators onto the garment meshes and hair strands. We then transfer the motion onto 3D Gaussians through carefully designed mechanisms for each body part. As a result, our synthesized avatars have vivid texture and realistic dynamic motion. To the best of our knowledge, our method is the first to produce highly realistic, fully simulation-ready 3D avatars, surpassing the capabilities of current approaches. Project page: https://research.nvidia.com/labs/dair/simavatar/

We present a novel contact model, termed Offset Geometric Contact (OGC), for guaranteed penetration-free simulation of codimensional objects with minimal computational overhead. Our method is based on constructing a volumetric shape by offsetting each face along its normal direction, ensuring orthogonal contact forces, thus allows large contact radius without artifacts. We compute vertex-specific displacement bounds to guarantee penetration-free simulation, which improves convergence and avoids the need for expensive continuous collision detection. Our method relies solely on massively parallel local operations, avoiding global synchronization and enabling efficient GPU implementation. Experiments demonstrate real-time, large-scale simulations with performance more than two orders of magnitude faster than prior methods while maintaining consistent computational budgets.

Various heuristic objectives for modeling hand-object interaction have been proposed in past work. However, due to the lack of a cohesive framework, these objectives often possess a narrow scope of applicability and are limited by their efficiency or accuracy. In this paper, we propose HANDYPRIORS, a unified and general pipeline for pose estimation in human-object interaction scenes by leveraging recent advances in differentiable physics and rendering. Our approach employs rendering priors to align with input images and segmentation masks along with physics priors to mitigate penetration and relative-sliding across frames. Furthermore, we present two alternatives for hand and object pose estimation. The optimization-based pose estimation achieves higher accuracy, while the filtering-based tracking, which utilizes the differentiable priors as dynamics and observation models, executes faster. We demonstrate that HANDYPRIORS attains comparable or superior results in the pose estimation task, and that the differentiable physics module can predict contact information for pose refinement. We also show that our approach generalizes to perception tasks, including robotic hand manipulation and human-object pose estimation in the wild.

Multi-finger grasping relies on high quality training data, which is hard to obtain: human data is hard to transfer and synthetic data relies on simplifying assumptions that reduce grasp quality. By making grasp simulation differentiable, and contact dynamics amenable to gradient-based optimization, we accelerate the search for high-quality grasps with fewer limiting assumptions. We present Grasp'D-1M: a large-scale dataset for multi-finger robotic grasping, synthesized with Fast-Grasp'D, a novel differentiable grasping simulator. Grasp'D-1M contains one million training examples for three robotic hands (three, four and five-fingered), each with multimodal visual inputs (RGB+depth+segmentation, available in mono and stereo). Grasp synthesis with Fast-Grasp'D is 10x faster than GraspIt! [1] and 20x faster than the prior Grasp'D differentiable simulator [2]. Generated grasps are more stable and contact-rich than GraspIt! grasps, regardless of the distance threshold used for contact generation. We validate the usefulness of our dataset by retraining an existing vision-based grasping pipeline [3] on Grasp'D-1M, and showing a dramatic increase in model performance, predicting grasps with 30% more contact, a 33% higher epsilon metric, and 35% lower simulated displacement. Additional details at fast-graspd.github.io.

Abstract In this work, we present a novel approach for calibrating material model parameters for soft body simulations using real data. We use a fully differentiable pipeline, combining a differentiable soft body simulator and differentiable depth rendering, which permits fast gradient‐based optimizations. Our method requires no data pre‐processing, and minimal experimental set‐up, as we directly minimize the L2‐norm between raw LIDAR scans and rendered simulation states. In essence, we provide the first marker‐free approach for calibrating a soft‐body simulator to match observed real‐world deformations. Our approach is inexpensive as it solely requires a consumer‐level LIDAR sensor compared to acquiring a professional marker‐based motion capture system. We investigate the effects of different material parameterizations and evaluate convergence for parameter optimization in both single and multi‐material scenarios of varying complexity. Finally, we show that our set‐up can be extended to optimize for dynamic behaviour as well.

Abstract We propose a hierarchical graph for learning physics and a novel way to handle obstacles. The finest level of the graph consist of the particles itself. Coarser levels consist of the cells of sparse grids with successively doubling cell sizes covering the volume occupied by the particles. The hierarchical structure allows for the information to propagate at great distance in a single message passing iteration. The novel obstacle handling allows the simulation to be obstacle aware without the need for ghost particles. We train the network to predict effective acceleration produced by multiple sub‐steps of 3D multi‐material material point method (MPM) simulation consisting of water, sand and snow with complex obstacles. Our network produces lower error, trains up to 7.0X faster and inferences up to 11.3X faster than [SGGP*20]. It is also, on average, about 3.7X faster compared to Taichi Elements simulation running on the same hardware in our tests.

Abstract The shape matching method is a popular approach to simulate deformable objects in interactive applications due to its stability and simplicity. An important feature is that there is no need for a mesh since the method works on arbitrary local groups within a set of particles. A major drawback of shape matching is the fact that it is geometrically motivated and not derived from physical principles which makes calibration difficult. The fact that the method does not conserve volume can yield visual artifacts, e.g. when a tire is compressed but does not bulge. In this paper we present a new meshless simulation method that is related to shape matching but derived from continuous constitutive models. Volume conservation and stiffness can be specified with physical parameters. Further, if the elements of a tetrahedral mesh are used as groups, our method perfectly reproduces FEM based simulations.

Tactile sensing is critical for robotic grasping and manipulation of objects under visual occlusion. However, in contrast to simulations of robot arms and cameras, current simulations of tactile sensors have limited accuracy, speed, and utility. In this work, we develop an efficient 3D finite element method (FEM) model of the SynTouch BioTac sensor using an open-access, GPU-based robotics simulator. Our simulations closely reproduce results from an experimentally-validated model in an industry-standard, CPU-based simulator, but at 75x the speed. We then learn latent representations for simulated BioTac deformations and real-world electrical output through self-supervision, as well as projections between the latent spaces using a small supervised dataset. Using these learned latent projections, we accurately synthesize real-world BioTac electrical output and estimate contact patches, both for unseen contact interactions. This work contributes an efficient, freely-accessible FEM model of the BioTac and comprises one of the first efforts to combine self-supervision, cross-modal transfer, and sim-to-real transfer for tactile sensors.

In computer graphics, soft body simulation is often used to animate soft tissue on characters or rubber like objects. Both are highly incompressible, however commonly used models such as co-rotational FEM, show significant volume loss, even under moderate strain. The Neo-Hookean model has recently become popular in graphics. It has superior volume conservation, recovers from inverted states, and does not require a polar decomposition. However, solvers for Neo-Hookean finite-element problems are typically based on Newton methods, which require energy Hessians, their Eigen-decomposition, and sophisticated linear solvers. In addition, minimizing the energy directly in this way does not accommodate modeling incompressible materials since it would require infinitely stiff forces. In this paper we present a constraint-based model of the Neo-Hookean energy. By decomposing the energy into deviatoric (distortional), and hydrostatic (volume preserving) constraints, we can apply iterative constrained-optimization methods that require only first-order gradients. We compare our constraint-based formulation to state-of-the-art force-based solvers and show that our method is often an order of magnitude more efficient for stiff volume preserving materials.

Abstract We introduce a triangle mesh based convolutional neural network. The proposed network structure can be used for problems where input and/or output are defined on a manifold triangle mesh with or without boundary. We demonstrate its applications in cloth upsampling, adding back details to Principal Component Analysis (PCA) compressed cloth, regressing clothing deformation from character poses, and regressing hand skin deformation from bones' joint angles. The data used for training in this work are generated from high resolution extended position based dynamics (XPBD) physics simulations with small time steps and high iteration counts and from an offline FEM simulator, but it can come from other sources. The inference time of our prototype implementation, depending on the mesh resolution and the network size, can provide between 4 to 134 times faster than a GPU based simulator. The inference also only needs to be done for meshes currently visible by the camera.

Abstract Inspired by frictional behaviour that is observed when sliding matchsticks against one another at different angles, we propose a phenomenological anisotropic friction model for structured surfaces. Our model interpolates isotropic and anisotropic elliptical Coulomb friction parameters for a pair of surfaces with perpendicular and parallel structure directions (e.g. the wood grain direction). We view our model as a special case of an abstract friction model that produces a cone based on state information, specifically the relationship between structure directions. We show how our model can be integrated into LCP and NCP‐based simulators using different solvers with both explicit and fully implicit time‐integration. The focus of our work is on symmetric friction cones, and we therefore demonstrate a variety of simulation scenarios where the friction structure directions play an important part in the resulting motions. Consequently, authoring of friction using our model is intuitive and we demonstrate that our model is compatible with standard authoring practices, such as texture mapping.

Abstract The “procedural” approach to animating ocean waves is the dominant algorithm for animating larger bodies of water in interactive applications as well as in off‐line productions — it provides high visual quality with a low computational demand. In this paper, we widen the applicability of procedural water wave animation with an extension that guarantees the satisfaction of boundary conditions imposed by terrain while still approximating physical wave behavior. In combination with a particle system that models wave breaking, foam, and spray, this allows us to naturally model waves interacting with beaches and rocks. Our system is able to animate waves at large scales at interactive frame rates on a commodity PC.

Abstract We examine the relationship between primal, or force‐based, and dual, or constraint‐based formulations of dynamics. Variational frameworks such as Projective Dynamics have proved popular for deformable simulation, however they have not been adopted for contact‐rich scenarios such as rigid body simulation. We propose a new preconditioned frictional contact solver that is compatible with existing primal optimization methods, and competitive with complementarity‐based approaches. Our relaxed primal model generates improved contact force distributions when compared to dual methods, and has the advantage of being differentiable, making it well‐suited for trajectory optimization. We derive both primal and dual methods from a common variational point of view, and present a comprehensive numerical analysis of both methods with respect to conditioning. We demonstrate our method on scenarios including rigid body contact, deformable simulation, and robotic manipulation.

Abstract We present a rigid body simulation method that can resolve small temporal and spatial details by using a quasi explicit integration scheme that is unconditionally stable. Traditional rigid body simulators linearize constraints because they operate on the velocity level or solve the equations of motion implicitly thereby freezing the constraint directions for multiple iterations. Our method always works with the most recent constraint directions. This allows us to trace high speed motion of objects colliding against curved geometry, to reduce the number of constraints, to increase the robustness of the simulation, and to simplify the formulation of the solver. In this paper we provide all the details to implement a fully fledged rigid body solver that handles contacts, a variety of joint types and the interaction with soft objects.

We consider the problem of transferring policies to the real world by training on a distribution of simulated scenarios. Rather than manually tuning the randomization of simulations, we adapt the simulation parameter distribution using a few real world roll-outs interleaved with policy training. In doing so, we are able to change the distribution of simulations to improve the policy transfer by matching the policy behavior in simulation and the real world. We show that policies trained with our method are able to reliably transfer to different robots in two real world tasks: swing-peg-in-hole and opening a cabinet drawer. The video of our experiments can be found at https://sites.google.com/view/simopt.

In this work we present a framework that is capable of accurately representing soft robotic actuators in a multiphysics environment in real-time. We propose a constraint-based dynamics model of a 1-dimensional pneumatic soft actuator that accounts for internal pressure forces, as well as the effect of actuator latency and damping under inflation and deflation and demonstrate its accuracy a full soft robotic snake with the composition of multiple 1D actuators. We verify our model's accuracy in static deformation and dynamic locomotion open-loop control experiments. To achieve real-time performance we leverage the parallel computation power of GPUs to allow interactive control and feedback.

We present a framework for the simulation of rigid and deformable bodies in the presence of contact and friction. Our method is based on a non-smooth Newton iteration that solves the underlying nonlinear complementarity problems (NCPs) directly. This approach allows us to support nonlinear dynamics models, including hyperelastic deformable bodies and articulated rigid mechanisms, coupled through a smooth isotropic friction model. The fixed-point nature of our method means it requires only the solution of a symmetric linear system as a building block. We propose a new complementarity preconditioner for NCP functions that improves convergence, and we develop an efficient GPU-based solver based on the conjugate residual (CR) method that is suitable for interactive simulations. We show how to improve robustness using a new geometric stiffness approximation and evaluate our method’s performance on a number of robotics simulation scenarios, including dexterous manipulation and training using reinforcement learning.

We introduce a deep reinforcement learning method that learns to control articulated humanoid bodies to imitate given target motions closely when simulated in a physics simulator. The target motion, which may not have been seen by the agent and can be noisy, is supplied at runtime. Our method can recover balance from moderate external disturbances and keep imitating the target motion. When subjected to large disturbances that cause the humanoid to fall down, our method can control the character to get up and recover to track the motion. Our method is trained to imitate the mocap clips from the CMU motion capture database and a number of other publicly available databases. We use a state-of-the-art deep reinforcement learning algorithm to learn to dynamically control the gain of PD controllers, whose target angles are derived from the mocap clip and to apply corrective torques with the goal of imitating the provided motion clip as closely as possible. Both the simulation and the learning algorithms are parallelized and run on the GPU. We demonstrate that the proposed method can control the character to imitate a wide variety of motions such as running, walking, dancing, jumping, kicking, punching, standing up, and so on.

Abstract Robustly and efficiently simulating cables and ropes that are part of a larger system such as cable driven machines, cable cars or tendons in a human or robot is a challenging task. To be able to adapt to the environment, cables are typically modeled as a large number of small segments that are connected via joints. The two main difficulties with this approach are to satisfy the inextensibility constraint and to handle the typically large mass ratio between the small segments and the larger objects they connect. In this paper we present a new approach which solves these problems in a simple and effective way. Our method is based on the idea to simulate the effect of the cables instead of the cables themselves. To this end we propose a new special type of distance constraint we call cable joint that changes both its attachment points and its rest length dynamically. A cable connecting a series of objects is then modeled as a sequence of cable joints which reduces the complexity of the simulation from the order of the number of segments to just the number of connected objects. This makes simulations both faster and more robust as we will demonstrate on a variety of examples.

The current state of the art in real-time two-dimensional water wave simulation requires developers to choose between efficient Fourier-based methods, which lack interactions with moving obstacles, and finite-difference or finite element methods, which handle environmental interactions but are significantly more expensive. This paper attempts to bridge this long-standing gap between complexity and performance, by proposing a new wave simulation method that can faithfully simulate wave interactions with moving obstacles in real time while simultaneously preserving minute details and accommodating very large simulation domains. Previous methods for simulating 2D water waves directly compute the change in height of the water surface, a strategy which imposes limitations based on the CFL condition (fast moving waves require small time steps) and Nyquist's limit (small wave details require closely-spaced simulation variables). This paper proposes a novel wavelet transformation that discretizes the liquid motion in terms of amplitude-like functions that vary over space, frequency, and direction , effectively generalizing Fourier-based methods to handle local interactions. Because these new variables change much more slowly over space than the original water height function, our change of variables drastically reduces the limitations of the CFL condition and Nyquist limit, allowing us to simulate highly detailed water waves at very large visual resolutions. Our discretization is amenable to fast summation and easy to parallelize. We also present basic extensions like pre-computed wave paths and two-way solid fluid coupling. Finally, we argue that our discretization provides a convenient set of variables for artistic manipulation, which we illustrate with a novel wave-painting interface.

We address the long-standing problem of iteration count and time step dependent constraint stiffness in position-based dynamics (PBD). We introduce a simple extension to PBD that allows it to accurately and efficiently simulate arbitrary elastic and dissipative energy potentials in an implicit manner. In addition, our method provides constraint force estimates, making it applicable to a wider range of applications, such those requiring haptic user-feedback. We compare our algorithm to more expensive non-linear solvers and find it produces visually similar results while maintaining the simplicity and robustness of the PBD method.

We present a novel algorithm to extract the rotational part of an arbitrary 3 X 3 matrix. This problem lies at the core of two popular simulation methods in computer graphics, the co-rotational Finite Element Method and Shape Matching techniques. In contrast to the traditional method based on polar decomposition, degenerate configurations and inversions are handled robustly and do not have to be treated in a special way. In addition, our method can be implemented with only a few lines of code without branches which makes it particularly well suited for GPU-based applications. We demonstrate the robustness, coherence and efficiency of our method by comparing it to stabilized polar decomposition in several simulation scenarios.

In computer graphics, simulated objects typically have two or three different representations, a visual mesh, a simulation mesh and a collection of convex shapes for collision handling. Using multiple representations requires skilled authoring and complicates object handing at run time. It can also produce visual artifacts such as a mismatch of collision behavior and visual appearance. The reason for using multiple representation has been performance restrictions in real time environments. However, for virtual worlds, we believe that the ultimate goal must be WYSIWYS -- what you see is what you simulate, what you can manipulate, what you can touch.

We present a novel explicit surface tracking method. Its main advantage over existing approaches is the fact that it is both completely grid-free and fast which makes it ideal for the use in large unbounded domains. A further advantage is that its running time is less sensitive to temporal variations of the input mesh than existing approaches. In terms of performance, the method provides a good trade-off point between speed and quality. The main idea behind our approach to handle topological changes is to delete all overlapping triangles and to fill or join the resulting holes in a robust and efficient way while guaranteeing that the output mesh is both manifold and without boundary. We demonstrate the flexibility, speed and quality of our method in various applications such as Eulerian and Lagrangian liquid simulations and the simulation of solids under large plastic deformations.

We propose a new method for both collision detection and collision response geared towards handling complex deformable objects in close contact. Our method does not miss collision events between time steps and solves the challenging problem of untangling automatically and robustly. It is conceptually simple and straight forward to parallelize due to the regularity of the algorithm. The main idea is to tessellate the air between objects once before the simulation and by considering one unilateral constraint per element that prevents its inversion during the simulation. If large relative rotations and translations are present in the simulation, an additional dynamic mesh optimization step is needed to prevent mesh locking. This step is fast in 2D and allows the simulation of arbitrary scenes. Because mesh optimization is expensive in 3D, however, the method is best suited for the subclass of 3D scenarios in which relative motions are limited. This subclass contains two important problems, namely the simulation of multi-layered clothing and tissue on animated characters.

Abstract The dynamic simulation of mechanical effects has a long history in computer graphics. The classical methods in this field discretize Newton's second law in a variety of Lagrangian or Eulerian ways, and formulate forces appropriate for each mechanical effect: joints for rigid bodies; stretching, shearing or bending for deformable bodies and pressure, or viscosity for fluids, to mention just a few. In the last years, the class of position‐based methods has become popular in the graphics community. These kinds of methods are fast, stable and controllable which make them well‐suited for use in interactive environments. Position‐based methods are not as accurate as force‐based methods in general but they provide visual plausibility. Therefore, the main application areas of these approaches are virtual reality, computer games and special effects in movies. This state‐of‐the‐art report covers the large variety of position‐based methods that were developed in the field of physically based simulation. We will introduce the concept of position‐based dynamics, present dynamic simulation based on shape matching and discuss data‐driven upsampling approaches. Furthermore, we will present several applications for these methods.

We present a unified dynamics framework for real-time visual effects. Using particles connected by constraints as our fundamental building block allows us to treat contact and collisions in a unified manner, and we show how this representation is flexible enough to model gases, liquids, deformable solids, rigid bodies and cloth with two-way interactions. We address some common problems with traditional particle-based methods and describe a parallel constraint solver based on position-based dynamics that is efficient enough for real-time applications.

In fluid simulation, enforcing incompressibility is crucial for realism; it is also computationally expensive. Recent work has improved efficiency, but still requires time-steps that are impractical for real-time applications. In this work we present an iterative density solver integrated into the Position Based Dynamics framework (PBD). By formulating and solving a set of positional constraints that enforce constant density, our method allows similar incompressibility and convergence to modern smoothed particle hydro-dynamic (SPH) solvers, but inherits the stability of the geometric, position based dynamics method, allowing large time steps suitable for real-time applications. We incorporate an artificial pressure term that improves particle distribution, creates surface tension, and lowers the neighborhood requirements of traditional SPH. Finally, we address the issue of energy loss by applying vorticity confinement as a velocity post process.