zombie-rs

Grid-free, no_std Monte Carlo geometry processing kernel in Rust inspired by Sawhney & Crane’s “Monte Carlo Geometry Processing: A Grid-Free Approach to PDE-Based Methods on Volumetric Domains”. The crate mirrors the spirit of the original C++ reference implementation zombie while embracing Rust’s safety and composability.

Current snapshot

• Walk-on-Spheres estimators for Laplace and (screened) Poisson Dirichlet problems on mesh and SDF domains.
• Observer hooks for per-walk telemetry and gradient estimators that reuse the same interface.
• Stateless solver façade (Solver, WalkBudget) for ergonomic usage in no_std environments.
• ASCII PLY dump format: vertices in walk order, RGB encodes role (start = cyan, steps = white, boundary hit = green, max-step exit = red).

Workspace layout

• crates/zombie_rs/ — core library modules and examples.
• crates/heat_probe/ — interactive heat-probe viewer (egui + eframe).

Quick start

cargo run -p heat_probe

Use the side panel to adjust slice height, resolution, and walk parameters while the image converges progressively in the centre panel. The viewer streams the Monte Carlo estimates directly from the kernel and can be paused/restarted by tweaking any control.

Why Monte Carlo geometry processing?

WoS replaces volumetric grids with random walks in the continuous domain, yielding unbiased solutions with controllable variance and trivial parallelisation. This workspace provides a Rust-first, safe foundation for research and application-level experimentation.