Recommendations to improve simulation performance in large building

Im trying to simulate a simple diff-drive robot (basically the one built here: Making a Mobile Robot with ROS - Full list of tutorials | Articulated Robotics ) in a large building (see image below).

Im using the gpu_lidar sensor as well as the camera sensor with gazebo fortress.

The simulation ran mostly fine in a small test-building with 3 simple rooms, that was built using the old gazebo-classic building editor.

But because that editor is quite limited and generates models that dont load out of the box in gazebo fortress. I looked for alternatives and stumbled on, wich can output collada models.

I created the following building-model (image created from the modelling software, not gazebo):

Using this model, the simulation now runs with a real time factor of about 35% (headless).

I’m playing around with the different parameters to get to an acceptable simulation rate (>= 90% is not expecting too much right?).

I can improve the RTF to about 50% by disabling shadows, but other than that I can only improve the performance by lowering the sensor resolutions (ie: lowering camera-sensor FPS or LiDAR update-rate).

Are there any other things I can do to improve the simulation performance that don’t result in worse sensor data?

Or maybe from another perspective: how would you approach modeling a building of the size shown in the image above?
I have no need for fancy textures, lighting or physics. Just lots of rooms with flat walls (and eventually an elevator, but thats for the future).
What software would you use to create the 3D model of the building? What gazebo-parameters would you focus on (compared to a smaller building)

You’re looking for the feature called Levels. You can partition the world to smaller pieces and load and unload them as they become relevant.

Thanks. Yes this looks promising.

Can you recommend a software to either edit (slice into levels) existing collada files, or model and create new buildings using levels?