arXiv subjects

Wenzhi Guo

Publications and source records attributed to Wenzhi Guo.

5 recordsLinked to original sources

arXiv2026

Rigorously justified local time stepping in the unified gas-kinetic wave-particle method for steady multiscale flow simulation

Local time stepping (LTS) can accelerate convergence to steady states in kinetic simulations with large variations in the local time steps across the computational domain. When neighboring cells advance with unequal time steps, the time-averaged particle flux must be balanced across their common interface. For particle-based or hybrid wave-particle methods under finite volume method (FVM) framework, we rigorously establish a sufficient condition for time-averaged interfacial particle-flux balance: fixed positive local time steps together with proportional particle-mass scaling. When a particle crosses from cell $L$ to cell $R$, its mass is scaled by $Δt_R/Δt_L$. In the unified gas-kinetic wave-particle (UGKWP) implementation, the same ratio is applied to the remaining free-transport time of the crossing particle. LTS also affects the wave-particle decomposition and the time integration of the wave fluxes in UGKWP. The wave-particle decomposition in cell $i$ is determined by the local ratio $Δt_i/τ_i$, which better reflects the local relation between the observation scale and relaxation time for multiscale cases. The equilibrium and analytic free transport wave fluxes are integrated over and normalized by the corresponding cell-side time steps to obtain the interfacial time-averaged wave fluxes. The UGKWP-LTS method is used to simulate the hypersonic flow past a cylinder at $\mathrm{Kn}=0.01$ and $0.1$, and a flat plate at $\mathrm{Kn}=0.0169$. In all three cases, the surface quantities obtained with UGKWP-LTS agree well with the reference data. Relative to global time stepping (GTS), UGKWP-LTS achieves step-count speedups of $6.6\times$, $3.8\times$, and $20\times$ for the three cases, respectively. The corresponding wall-clock speedups are $7.1\times$, $4.5\times$, and approximately $21\times$.

physics.flu-dyn↗

arXiv2026

Surrogate-Based Aerodynamic Shape Optimization in Multiscale Flows via the Implicit Unified Gas-Kinetic Scheme

While hypersonic glide vehicles such as the HTV-2 continue to be a focal point in aerospace research, their aerodynamic characteristics in complex near-space environments are not yet fully understood. Because traditional continuum assumptions fail to accurately capture multiscale flow features across varying rarefied altitudes, this study investigates the aerodynamic shape optimization of an HTV-2-type aircraft across multiple flow regimes. An automated optimization framework is developed by coupling surrogate-based optimization (SBO) with the implicit unified gas-kinetic scheme (IUGKS). To ensure relevance to practical engineering requirements, both volumetric and center-of-pressure constraints are incorporated into the optimization process. The resulting optimized configurations are subsequently validated through high-fidelity computations, detailed flow-field evaluations, and global sensitivity analyses. Under volumetric constraints, the optimized lift-to-drag ratio ($L/D$) increases significantly at altitudes ranging from 70 km to 100 km. The optimal aerodynamic strategy is shown to shift with altitude: at 70 km, reducing the windward radius ($R_1$) weakens the oblique shock wave, whereas at highly rarefied altitudes, reducing the leeward radius ($R_3$) enhances the expansion wave. Correspondingly, sensitivity analyses confirm that as flow rarefaction increases, aerodynamic dominance shifts toward $R_3$. Furthermore, reducing the wingtip bluntness ($R_2$)yields consistent aerodynamic benefits across the entire flight envelope, ultimately driving the optimized geometries toward a flatter and more slender profile.

physics.flu-dyn↗

arXiv2026

PocketGS: On-Device Training of 3D Gaussian Splatting for High Perceptual Modeling

While 3D Gaussian Splatting (3DGS) enables real-time rendering, its training demands workstation-level compute and memory, making mobile deployment impractical under minute-scale time budgets and limited peak memory. We present PocketGS, a mobile scene modeling paradigm that enables on-device 3DGS training under these tightly coupled constraints while preserving high-fidelity reconstruction. PocketGS resolves the fundamental tension between training efficiency, memory compactness, and modeling quality through three co-designed operators: $\mathcal{G}$ builds geometry-faithful point-cloud priors; $\mathcal{I}$ injects local surface statistics to seed anisotropic Gaussians, thereby reducing early conditioning gaps; and $\mathcal{T}$ unrolls alpha compositing with cached intermediates and index-mapped gradient scattering for stable mobile backpropagation. Extensive experiments demonstrate that PocketGS outperforms the powerful mainstream workstation 3DGS baseline under mobile budgets, delivering high-quality reconstructions and enabling a fully on-device, practical capture-to-rendering workflow.

cs.CV↗

arXiv2025

Accurate and Complete Surface Reconstruction from 3D Gaussians via Direct SDF Learning

3D Gaussian Splatting (3DGS) has recently emerged as a powerful paradigm for photorealistic view synthesis, representing scenes with spatially distributed Gaussian primitives. While highly effective for rendering, achieving accurate and complete surface reconstruction remains challenging due to the unstructured nature of the representation and the absence of explicit geometric supervision. In this work, we propose DiGS, a unified framework that embeds Signed Distance Field (SDF) learning directly into the 3DGS pipeline, thereby enforcing strong and interpretable surface priors. By associating each Gaussian with a learnable SDF value, DiGS explicitly aligns primitives with underlying geometry and improves cross-view consistency. To further ensure dense and coherent coverage, we design a geometry-guided grid growth strategy that adaptively distributes Gaussians along geometry-consistent regions under a multi-scale hierarchy. Extensive experiments on standard benchmarks, including DTU, Mip-NeRF 360, and Tanks& Temples, demonstrate that DiGS consistently improves reconstruction accuracy and completeness while retaining high rendering fidelity.

cs.CV↗

arXiv2024

NeuV-SLAM: Fast Neural Multiresolution Voxel Optimization for RGBD Dense SLAM

We introduce NeuV-SLAM, a novel dense simultaneous localization and mapping pipeline based on neural multiresolution voxels, characterized by ultra-fast convergence and incremental expansion capabilities. This pipeline utilizes RGBD images as input to construct multiresolution neural voxels, achieving rapid convergence while maintaining robust incremental scene reconstruction and camera tracking. Central to our methodology is to propose a novel implicit representation, termed VDF that combines the implementation of neural signed distance field (SDF) voxels with an SDF activation strategy. This approach entails the direct optimization of color features and SDF values anchored within the voxels, substantially enhancing the rate of scene convergence. To ensure the acquisition of clear edge delineation, SDF activation is designed, which maintains exemplary scene representation fidelity even under constraints of voxel resolution. Furthermore, in pursuit of advancing rapid incremental expansion with low computational overhead, we developed hashMV, a novel hash-based multiresolution voxel management structure. This architecture is complemented by a strategically designed voxel generation technique that synergizes with a two-dimensional scene prior. Our empirical evaluations, conducted on the Replica and ScanNet Datasets, substantiate NeuV-SLAM's exceptional efficacy in terms of convergence speed, tracking accuracy, scene reconstruction, and rendering quality.

cs.CV↗