Monte Carlo Light Simulation in Voxel Models

As part of ongoing research with CITA's Eco-Metabolistic Architecture unit at the Royal Danish Academy, this project develops a voxel-based workflow for simulation and design. Leveraging the parallel processing power of modern GPUs, we bring together modeling, raytracing, simulation, and agent-based automata in a single holistic model. The growth of cyanobacteria (Fremyella Diplosiphon aka Tolypothrix sp. PCC 7601) is modeled using cellular automata while spatial light intensity maps are integrated on the fly using techniques of general purpose GPU programming.

The simulation treats light as individual particles, tracking millions of virtual photons each frame, and achieving realtime speeds (> 30-60 fps). The medium is modeled as voxels and carries complex information including material and density gradients, and variable optical properties such as the scattering coefficient and the index of refraction. The voxel grid supports each step of the pipeline from light simulation to bacterial growth simulation to geometry generation, bypassing costly remeshing steps. These approximations allow these models to run at far faster speeds than many more exact multiphysics solvers, with nevertheless comparable results. Enabling instant design feedback brings these tools into the world of design, and targets architectural workflows seeking to leverage the power of scientific simulation practically and flexibly.

Geometries can easily be transfered between traditional CAD and 3d modeling software and the simulation engine.

Cellular automata and other agent-based models are naturally suited to voxel grids.

Light Simulation outputs a volumetric intensity "heatmap" suitable for analysis or further simulation.

Credits:
  • Martin Tamke
  • Mette Ramsgaard-Thomsen
  • Michael Kühl
  • Shahriar Akbari
  • Swathi Murthy