physics engine integration

This work addresses the challenge of accurately and physically consistently capturing sharp gradients such as shock waves in hypersonic flow fields, which are poorly resolved by conventional reduced-order models or neural surrogates. To this end, the authors propose an end-to-end fully GPU-accelerated workflow that leverages the differentiable high-fidelity solver JAX-Fluids for efficient data generation and introduces a residual-driven, physics-aware refinement mechanism. The resulting neural surrogate is trained using only mesh coordinates and input parameters, significantly reducing residuals of the governing equations while improving the physical fidelity of predicted shock locations and strengths. The method demonstrates strong generalization and reliability even under out-of-distribution operating conditions.

Existing CPS co-simulation tools suffer from limited portability, modularity, and automation. To address these limitations, this paper proposes a Python-based programmable co-simulation framework. The framework enables declarative orchestration and runtime dynamic substitution of multi-fidelity heterogeneous components—breaking away from conventional static configuration paradigms. It adopts a componentized architecture, supports distributed communication via ZeroMQ and ROS, and provides standardized adaptation interfaces for third-party platforms (e.g., PX4), thereby enabling cross-platform, reconfigurable co-simulation. Its core innovation is the first-ever declarative component orchestration mechanism, which significantly enhances simulation system reusability, reproducibility, and development efficiency. The framework is validated through co-simulation of unmanned aerial vehicles and autonomous controllers, demonstrating its flexibility and practicality. This work establishes a novel paradigm for CPS benchmark construction and automated evaluation.

This work addresses the challenges of high-fidelity simulation in fusion energy systems—specifically, geometric modeling, multiphysics coupling, and the integration of particle and continuum methods—by proposing a unified framework that seamlessly combines commercial CAE software with existing fusion codes. The framework enables accurate representation of complex geometries, automatic generation of unstructured meshes, and efficient coupling between particle transport and continuum solvers. The resulting simulation workflow significantly enhances geometric fidelity for critical components and strengthens capabilities in multiscale, multiphysics co-simulation, while maintaining strong scalability and computational efficiency.

This study addresses the low physical fidelity of generative video models—manifested as artifacts such as jittering and interpenetration—by proposing a physics-aware enhancement method grounded in synthetic video. Methodologically, it employs a differentiable rendering pipeline to generate physically consistent synthetic videos, establishes a physics-perceptive data filtering mechanism, and introduces cross-domain feature alignment coupled with adversarial physical consistency regularization—enabling physics realism transfer without differentiable simulation or explicit physical modeling. This work provides the first empirical evidence that synthetic video can substantially improve physical fidelity in video generation. Evaluated on three physics-sensitive tasks—rigid-body collisions, fluid motion, and pendulum dynamics—the approach reduces physical violation rates significantly, achieving an average 37.2% improvement in physical plausibility, validated jointly by user studies and automated physical violation detection.

Existing 3D generative models produce high-fidelity geometries but fail to guarantee engineering-relevant physical properties—such as automotive aerodynamic efficiency—due to the absence of explicit physical modeling. To address this, we propose a physics-guided 3D shape generation framework. Our method introduces a physics-aware flow matching model integrated with an alternating update mechanism that embeds physical constraints directly into the generative process. Additionally, we design a Shape-Physics joint variational autoencoder (SP-VAE) that unifies flow matching, physics-based regularization, and latent-space physics fine-tuning. Extensive experiments across three benchmarks demonstrate that our approach significantly outperforms state-of-the-art methods in both physical plausibility and visual fidelity. To the best of our knowledge, this is the first work to achieve end-to-end differentiable 3D structural synthesis explicitly driven by physical laws.

This work addresses the challenge of constructing effective test oracles for Functional Mock-up Unit (FMU) simulation models, which lack explicit expected outputs and thus hinder the application of traditional testing methods. Furthermore, existing approaches to extracting metamorphic relations rely heavily on manual effort, resulting in low efficiency and susceptibility to human error. To overcome these limitations, the paper proposes a novel large language model (LLM)-based multi-agent workflow that, for the first time, integrates LLMs with multi-agent collaboration to automatically derive requirements from functional and interface specifications and generate structured Given-When-Then metamorphic relations. These relations drive the automated generation of metamorphic test cases and consistency validation for FMUs. Experimental evaluation on an oil cooling system FMU demonstrates that the approach significantly reduces manual intervention while enhancing the systematicity and efficiency of dynamic simulation model verification.

This work addresses the challenge of efficiently integrating dynamic 3D reconstruction with physics simulation, which is hindered by the complexity of collision detection under changing mesh topologies. The authors propose a dual-representation framework that employs a fixed-topology mesh to enable efficient physical simulation while leveraging Gaussian splatting for high-quality rendering. To handle topological changes, they introduce strategies including vertex buffer updates, temporal correspondence tracking, and stencil projection. Their systematic evaluation—the first of its kind—demonstrates a 4.65× speedup in simulation compared to variable-topology baselines, albeit at the cost of a 65–80% reduction in geometric fidelity during topology transitions. These findings reveal a fundamental trade-off between high-fidelity reconstruction and physics-compatible mesh topologies.

This work addresses the frequent mismatch between user-specified physical intent and the actual behavior of multiphysics simulation code generated by large language models, often due to erroneous implementations of partial differential equations (PDEs). To bridge this gap, we propose a PDE-structure-based intent verification method that deterministically reconstructs the governing equations implicitly encoded in the generated code and compares them against the user’s intended PDEs, enabling semantic correctness validation and iterative refinement. We introduce, for the first time, a formal metric termed the Intent Fidelity Score (IFS) to quantify alignment with physical intent, establish a PDE-driven feedback loop, and demonstrate compatibility with major PDE frameworks including MOOSE, FEniCS, and FreeFEM. Evaluated on 220 cases in MooseBench, our approach substantially improves IFS—by 0.22–0.41 on challenging instances with initial IFS < 0.7—while audits reveal that execution-only repair strategies still yield physically incorrect results in 39–40% of cases.

This study addresses the high computational cost of finite element simulation in industrial electric motor design, which severely hinders design iteration and product development efficiency. For the first time in industrial-scale motor simulations, this work integrates fine-grained runtime performance profiling with customized acceleration strategies to systematically identify critical computational bottlenecks in both two- and three-dimensional simulations. By applying targeted optimizations to core algorithms, the proposed approach substantially reduces CPU time overhead and achieves end-to-end simulation acceleration. This advancement effectively shortens time-to-market and establishes an efficient computational paradigm for high-fidelity motor design.

This work proposes the first end-to-end embodied AI system fully built upon the AMD ROCm ecosystem, addressing the prevailing reliance of vision-language-action (VLA) models on CUDA. By integrating the SmolVLA model, 3D Gaussian Splatting, and the Genesis physics engine, the system enables full-stack acceleration—from data-center training and photorealistic simulation rendering to edge inference on Ryzen AI platforms—within the ROCm+PyTorch framework. It establishes a complete Real2Sim2Real loop and demonstrates successful deployment of language-guided semantic manipulation policies on a Franka robotic arm. Furthermore, large-scale reinforcement learning is validated across quadrupedal and humanoid robot platforms, confirming the feasibility and efficiency of a pure AMD software stack for complex embodied intelligence tasks.