Static meshes optimised?

[chebob69]

Quickest, and probably simplest question ever:

Is there any benefit in creating collision rules for various static meshes that you've positioned overlapping each other? E.g a tall box representing a building that intersects the ground.
Or does their nature mean that bepu knows they won't interact with each other?

Thanks

[Norbo]

Static/kinematic objects are not tested against each other by default. The mechanism behind it is actually collision rules; static and kinematic objects belong to a different CollisionGroup by default that only collides with dynamic objects (CollisionRules.DefaultKinematicCollisionGroup versus CollisionRules.DefaultDynamicCollisionGroup).

That said, there still remains a tiny cost of actually evaluating those collision rules and realizing the pair can be skipped. It's extremely trivial compared to a full pair test, but you still may find in large simulations that avoiding it is useful. A StaticGroup containing the relevant static objects is a good option in such a case.

[chebob69]

Ahh, excellent. Haven't encountered the StaticGroup yet. From a look at the error messages, you have to give it a list of collidables for its constructor?
Can you not assign groups like you did with the collision group? What if you wanted to add more to it at runtime?

[Norbo]

StaticGroups are the static equivalent of compound collidables. Since a hierarchy is created for the set of contained objects, any additions or removals require a hierarchy update. Currently, the StaticGroup (and CompoundShape) do not support mutability. That's on the roadmap, though.