Octree Generation
When running an environment with a sonar sensor, octrees must be generated. These can often take up several gigabytes of storage, thus aren’t feasible as part of the downloaded package.
Upon startup, all octrees within the InitOctreeRange parameter are created,
then more are made as the agent moves throughout the environment. This can cause pauses
in the simulation the first time it is ran. A warning will appear onscreen about this
and can be disabled with the ShowWarning parameter. All subsequent simulation runs
will use the cached octrees and run much faster.
One option to avoid waiting times, is to run the simulation without the viewport, and let it generate octrees in the background. Here’s an example script that does exactly that, just change the scenario to the one that you would like to create octrees for,
import holoocean
import numpy as np
from tqdm import tqdm
command = np.array([0,0,0,0,-20,-20,-20,-20])
print("Building octrees...")
with holoocean.make("PierHarbor-HoveringImagingSonar", show_viewport=False) as env:
for i in tqdm(range(1000)):
env.act("auv0", command)
state = env.tick()
print("Finished Simulation")
When octrees are made, they are saved in the Package Installation Location in an octree folder. This will be as follows,
Platform |
Octree Save Location |
|---|---|
Linux |
|
Windows |
|
In this octree folder, there will be additional folders for each level name, and in those a folder for each octree size used. If files are being actively saved here it means that the simulation is still running and isn’t frozen.