Ray Tracing at City Scale

Important

This tutorial is still TODO, but you can find premise of it below.

The default path tracing solver, SBRPathTracer (solver="sbr"), discovers path candidates with a bounded population of shooting-and-bouncing rays instead of enumerating them, so its cost does not grow combinatorially with the reflection order or the number of primitives in the scene. This makes it possible to simulate higher-order reflections on city-scale scenes, like the one below, where an exhaustive search would be far too slow.

import differt.plotting as dplt
import jax.numpy as jnp
from differt.geometry import Mesh, Scene

Warp CUDA warning: Could not find or load the NVIDIA CUDA driver. GPU execution will not be available.

%%time

mesh_file = "bruxelles.obj"
mesh = Mesh.load_obj(mesh_file)
dplt.set_backend("plotly")  # Let's use the Plotly backend

tx = jnp.array([-40.0, 75, 30.0])
rx = jnp.array([+20.0, 108.034, 1.50])

scene = Scene(transmitters=tx, receivers=rx, mesh=mesh)

with dplt.reuse() as fig:
    scene.plot()
    paths = scene.trace_paths(order=range(6), solver="sbr")
    paths.plot()

fig

CPU times: user 9.17 s, sys: 448 ms, total: 9.61 s
Wall time: 7.52 s