multibody dynamics modeling

To address performance degradation in sim-to-real transfer for embodied intelligence—caused by modeling discrepancies in physics simulators—this paper presents the first systematic, three-dimensional evaluation of mainstream engines (e.g., PyBullet, MuJoCo, Isaac Gym) along physical fidelity, task adaptability, and hardware constraints. We integrate cutting-edge techniques—including world models and geometrically equivariant networks—to establish a comprehensive benchmark featuring multi-task datasets, unified evaluation metrics, and an open-source platform. Furthermore, we propose a task-aware simulator selection framework that quantifies trade-offs among accuracy, real-time capability, differentiability, and deployment compatibility for navigation and manipulation tasks. Our contributions include an open-source evaluation repository and practical guidelines, providing both theoretical foundations and engineering evidence to reduce real-world training costs and enhance transfer robustness.

This paper presents a systematic survey of state-of-the-art 3D modeling techniques for articulated objects, addressing the fundamental challenge of jointly modeling geometry—i.e., part structure and shape—and motion—i.e., dynamics and kinematic constraints. It establishes, for the first time, a unified taxonomy of co-design paradigms that integrate geometric and motion reasoning, rigorously defines task boundaries, and identifies core bottlenecks in generalization, physical plausibility, and cross-domain transfer. The survey comprehensively covers major technical approaches, including optimization-based methods, multi-view and single-image reconstruction, NeRFs, implicit neural representations, graph neural networks, and physics-based simulation. Building on this analysis, the authors propose the first structured classification framework for articulated object modeling, distilling seven key challenges and four concrete future research directions. This work serves as an authoritative benchmark and roadmap for researchers in computer vision, computer graphics, and robotics.

Existing simulation and embodied intelligence systems struggle to model fine-grained dynamical effects of articulated objects, such as frictional holding, positional sticking, and damped closure. This work proposes a structured three-channel field representation that explicitly captures conservative forces, dry friction, and damping along joint degrees of freedom, enabling inference and composition of interpretable dynamical primitives from vision-language inputs. For the first time, joint dynamics are formulated as a composable, differentiable function space compatible with physics-based simulation. By integrating shape-constrained piecewise cubic Hermite interpolation (PCHIP) with gradient-based optimization, the method achieves realistic and controllable modeling of complex mechanical behaviors. The framework provides a unified interface for dynamics inference, editing, and optimization, and will be accompanied by open-sourced code and example assets.

Long-term prediction of complex multi-body collisions in rigid-body dynamics remains challenging, with existing graph neural networks (GNNs) exhibiting significant limitations in modeling long-range interactions and generalizing across diverse geometric configurations and initial conditions. Method: We propose a novel physics-informed neural dynamics model that integrates physical priors with high-order topological structures. Specifically, we introduce simplicial complex embeddings into neural networks to enable topology-aware, physically constrained message passing, and jointly regularize the model using rigid-body dynamical equations and physics-informed neural networks (PINNs). Contribution/Results: Our approach achieves substantial improvements in long-horizon rollout accuracy and demonstrates strong generalization to unseen geometries and initial states. By embedding domain knowledge—both physical laws and topological relationships—into the architecture, the model yields interpretable, robust, and broadly applicable simulations for multi-body systems, establishing a generalizable and explainable paradigm for neural dynamical modeling.

This study addresses the limitations of traditional physics simulators in robotics—such as restricted expressiveness due to simplifying assumptions, high data costs, and difficulties in modeling complex physical interactions—by systematically reviewing video generation models as embodied world models. Integrating high-fidelity, multimodal-conditioned video synthesis with imitation learning, reinforcement learning, and visual planning frameworks, this work provides the first comprehensive analysis of their potential and limitations in tasks including action prediction, dynamics modeling, and policy evaluation. The review highlights breakthroughs in high-fidelity modeling of physical interactions while identifying key challenges in instruction following, physical consistency, and safety. These insights lay a theoretical foundation and outline future directions for replacing conventional simulators and enabling deployment in safety-critical scenarios.

Existing physics engines for robotics struggle to simultaneously ensure stable simulation, high-fidelity rigid contact modeling, and full differentiability with respect to states, actions, and system parameters. To address this, we propose Dojo—the first end-to-end differentiable physics engine designed specifically for robotics. Dojo uniquely integrates variational integrators with a second-order cone nonlinear complementarity problem (NCP) solver, guaranteeing energy and momentum conservation during contact and enabling smooth, analytic gradient computation across contact events. It further employs a customized primal-dual interior-point method for efficient implicit differentiation. Evaluated on motion planning, policy optimization, and system identification tasks, Dojo demonstrates significantly improved gradient accuracy and faster optimization convergence in challenging rigid-contact scenarios.

To address the low computational efficiency and poor accuracy of derivative computation for contact dynamics—particularly collisions and friction—in robotic simulation, which hinder convergence in reinforcement learning and trajectory optimization, this paper introduces the first unified analytical differentiation method for nonsmooth contact events. The method jointly exploits the sparsity of multibody systems and the intrinsic nonsmoothness of contact dynamics. Implemented in C++ and integrated into the Simple simulator, it supports arbitrary-degree-of-freedom rigid-body systems. On a 7-DOF manipulator and a 36-DOF humanoid robot, derivative evaluation takes only 5 μs and 95 μs, respectively—over 100× faster than state-of-the-art methods. This marks the first demonstration of microsecond-scale, high-accuracy, end-to-end differentiable robotic simulation.

This work addresses the critical influence of configuration space selection on constraint satisfaction accuracy in numerical simulations of fully constrained rigid body dynamics. From a geometric perspective, the study proposes a differential-algebraic equation (DAE) formulation and geometric integration scheme based on the Lie group SE(3). It demonstrates that when kinematic constraints correspond to subgroups of SE(3), these constraints can be preserved exactly over time. The approach elucidates the intrinsic relationship between SE(3) subgroup structures and lower-pair joints, establishing that employing SE(3) as the configuration space enables strict enforcement of constraints. This result provides both a theoretical foundation and numerical guarantees for high-fidelity simulation of rigid multibody systems.

Existing 3D datasets commonly lack the physical attributes necessary for stable and realistic interactions, necessitating extensive manual effort to construct simulation-ready articulated objects. This work introduces, for the first time, the concept of “interaction readiness” and establishes a quantifiable evaluation framework that decomposes its essential components. Building upon this foundation, the authors propose a method that combines multimodal perception fusion with closed-loop simulator optimization to automatically generate articulated objects with high-quality physical properties from incomplete 3D assets. By jointly reasoning over geometric, visual, and semantic cues to infer physical parameters and iteratively refining them through simulation feedback, the approach significantly enhances dynamic stability, interaction plausibility, and downstream policy learning performance across diverse manipulation tasks, while uncovering critical failure modes overlooked by conventional evaluation metrics.

Simulating large-scale articulated rigid-body systems remains challenging for conventional rigid-body solvers due to geometric nonlinearities and numerical stiffness. This work proposes a co-rotational framework based on Affine Body Dynamics (ABD) that decouples geometric nonlinearities through a linear kinematic mapping and projects high-dimensional body coordinates onto a dual space spanned by the minimal joint degrees of freedom. By combining implicit integration with KKT system solves, the method enforces exact constraint satisfaction and ensures physically accurate motion propagation. It supports diverse topologies—including chains, trees, closed loops, and irregular networks—and leverages pre-factorization of constant-coefficient matrices to achieve significant computational efficiency. The approach enables interactive simulation of systems comprising hundreds of thousands of rigid bodies on a single CPU core, maintaining high stability and accuracy even with large time steps.

This study addresses the challenge of accurately modeling the dynamics of flexible two-degree-of-freedom robotic arms, where rigid-body assumptions fail to capture link flexibility and unmodeled residual dynamics. To overcome this limitation, the authors propose a semi-parametric hybrid dynamical modeling framework that augments rigid-body dynamics with a Gaussian Mixture Model (GMM) to learn residual terms, while employing a purely data-driven kinematic regression as a baseline. This approach synergistically integrates physical priors with data-driven mechanisms, transcending the constraints of conventional fully parametric models in flexible systems. Experimental results on an open-source dataset demonstrate that the data-driven component—combined with regularization and least-squares estimation—significantly enhances torque prediction accuracy, thereby validating the efficacy and superiority of the proposed hybrid model.

Existing 3D generation methods often neglect physical properties or are confined to a single object category, failing to meet the demand for diverse and physically plausible assets in downstream simulation tasks. This work proposes PhysX-Omni, a unified framework that achieves joint generation of rigid, deformable, and articulated 3D objects with physical fidelity for the first time. Key innovations include a compression-free, high-resolution geometric representation tailored for vision-language models, the first universal simulation-ready 3D dataset—PhysXVerse—and PhysX-Bench, a comprehensive evaluation benchmark encompassing six-dimensional physical attributes. Experiments demonstrate that the proposed method excels on both conventional metrics and PhysX-Bench, significantly enhancing performance in downstream applications such as simulation scene generation and robot policy learning.