Warp-Geo: Differentiable Geometry Representation for Dynamic-Boundary Simulation and Shape Optimization

Warp-Geo is a differentiable GPU-accelerated framework designed for reconstructing signed distance fields (SDF) and evaluating surface normals from point clouds of complex 3D geometries. The framework addresses inverse problems and adaptive simulations that require efficient geometry representations frequently updated for gradient-based optimization. While applicable across multiple physics domains, the method is particularly relevant to shape optimization tasks in structural mechanics and fluid dynamics where boundary geometry plays a critical role in simulation accuracy.