Problem: baking Navigation Region does not provide the ability to exclude regions like roofs or inside of buildings, this mess up totally the agent navigation.
Adding collision box do not solve the problem because it creates the navigation on top and inside of it.
Am I missiong obvious settings or godot 3D navigation is not suitable ?
I just tried with 4.3.rc1 and I did not succeed. I checked “affect navigation mesh” to true, then adjusting the scale of the shape to the shape of my building (because I can’t use rod dot handle, dont see theme and when one is appearing I can’t click and drag it). Nothing seems to happen when I clear/bake navigation again
There is still something strange, for some buildings the NavigationObstacle3D yellow shape is not the same as definied in its .tscn.
And the navigation still happens on some area of the building where the shape is not the same
After getting it work by defining the shape with the packedvectors, the NavigationObstactle makes my fps dropping from 90 to 8. After deleting it it came back to 90 fps. So it is unusable
You dont need to delete the obstacle, you only need to disable the avoidance that is enabled by default.
The avoidance costs a lot of performance when the avoidance space needs to update for its agents and the obstacle shapes are numerous or more complex.
An obstacle with disabled avoidance costs nothing as all that is done with it is that it gets parsed with the navmesh baking. No baking, no cost, except for the loading at the begin.
An obstacle can not scale or rotate as it only has a position. The 3D gizmo always shows a full transform gui because it can not disable the parts that are not working because a Node3D has everything.
If you want to change the “scale” or “rotation” of an obstacle you need to change the vertices of the obstacle. The transform property of a Node3D is useless except for the position. An obstacle shape needs to be axis-aligned so it can not be rotated freely. In the navmesh baking process the obstacle shape will get rasterized on the voxel grid and every cell that is affected will be discarded, that is why it can remove internal geometry while other shapes cant. So the detail of the obstacle shape depends on the cell size resolution of the voxels.