AI Perception for large objects

I haven’t found anywhere in the docs that explicitly state this - but I have experimented and seen discussion threads to the effect that an AI sight stimuli source is always set to the center of a capsule for a Pawn/Character. I can also use the AI perception debug draw to see that for small objects, the ‘stimuli source center’ seems to be the center of the static-mesh.

So far so good.

My problem is that I can’t get my NPCs to see anything very large - like a building. I wonder if this is because that (presupposed) ‘stimuli center’ is hidden deeply within the large object and is getting occluded by the AI sight system? But I would have guessed that the AI sight would ignore all occluding geometry that is part of the actor that is being perceived?

If there’s some place in the docs that explain this, I’d love to see it - or hear from anyone doing AI perception against large structures.

My setup is a BP with a single very large static mesh and a stimuli-source - just as I have on the small objects - the only difference is the size of the mesh.

Thanks
Marcus

Hi, it should work, did you make sure that the distance between location of the object and what should perceive it is inside the sight perception range specified in the AI perception component?

Yes - I did check that.

I also have some large buildings being seen while others are not.

Even if I get this to work, the UE perception model isn’t really designed for large structures by placing the stimuli at the center. For a large structure, you’d want the perception to test for the closest bbox face to the viewer - not the center of the object.

I am looking into building my own perception system instead - I had to do so for sound anyway.