Implement bounding volumes for primitive shapes #11336
Conversation
Jondolf
added
C-Feature
I think we should split this out into a follow-up to avoid controversy. Personally, i think you probably want some sort of explicit projection to 2D before using this: the rotation is not always consistent. Alternatively, we could add 2D versions of the missing useful shapes (like capsules). |
|
Should be ready for review now. I fixed some docs, simplified line and plane stuff, and added basic tests for all of the bounding implementations. Most of the tests are for the trivial non-rotated case since computing the expected values for the rotated versions can be rather time-consuming, but at least the core functionality should now be tested. I think it'd be enough if we later added bounding volume examples (like in my earlier videos) where you could see the bounding volumes at all kinds of rotations and shapes and sizes. I also changed the |
|
Hmm... I'm not sure if we actually want to use the circumcircle/circumsphere for triangles, cones and conical frusta. The old approach of getting the AABB and computing the bounding circle for that might be better for more cases. For some cases, the circumcircle does give a tighter fit: (circumcircle) (bounding circle of AABB) But if the triangle has one long side and two shorter legs, the circumcircle puts the center very far away since it needs to pass through all vertices: (circumcircle) (bounding circle of AABB) I feel like the bounding circle of the AABB is better on average even if it doesn't give the tightest fit in all cases. It's also more consistent in that it has the same center as the AABB instead of some completely different center. Would it be good if I changed the triangle, cone and conical frustum implementations to use the AABB bounding circle/sphere approach, or am I perhaps missing some better alternative? I think the |
I think swapping it back is good: the consistency is valuable. I like the circumcircle helper still: it's a nice little algorithm. Having a slower method to compute an exact minimal bounding volume would also be useful for static cases. |
| /// The bounding sphere is not guaranteed to be the smallest possible. | ||
| #[inline(always)] | ||
| pub fn from_point_cloud(translation: Vec3, rotation: Quat, points: &[Vec3]) -> BoundingSphere { | ||
| let center = point_cloud_3d_center(points); |
There was a problem hiding this comment.
Okay so here I'm not super sure, but maybe instead of taking the geometric center, taking the min-max center (center of the AABB) might produce better results. (Same would apply to the 2D case I guess). It would have certain desirable properties (like being independent of points that are in the middle of the cloud), but I'm not sure it'd be better in real cases.
alice-i-cecile
added
the
S-Ready-For-Final-Review
# Objective Currently, the `Capsule` primitive is technically dimension-agnostic in that it implements both `Primitive2d` and `Primitive3d`. This seems good on paper, but it can often be useful to have separate 2D and 3D versions of primitives. For example, one might want a two-dimensional capsule mesh. We can't really implement both 2D and 3D meshing for the same type using the upcoming `Meshable` trait (see #11431). We also currently don't implement `Bounded2d` for `Capsule`, see #11336 (comment). Having 2D and 3D separate at a type level is more explicit, and also more consistent with the existing primitives, as there are no other types that implement both `Primitive2d` and `Primitive3d` at the same time. ## Solution Rename `Capsule` to `Capsule3d` and add `Capsule2d`. `Capsule2d` implements `Bounded2d`. For now, I went for `Capsule2d` for the sake of consistency and clarity. Mathematically the more accurate term would be `Stadium` or `Pill` (see [Wikipedia](https://en.wikipedia.org/wiki/Stadium_(geometry))), but those might be less obvious to game devs. For reference, Godot has [`CapsuleShape2D`](https://docs.godotengine.org/en/stable/classes/class_capsuleshape2d.html). I can rename it if others think the geometrically correct name is better though. --- ## Changelog - Renamed `Capsule` to `Capsule3d` - Added `Capsule2d` with `Bounded2d` implemented --------- Co-authored-by: Alice Cecile <[email protected]>
# Objective Currently, the `Capsule` primitive is technically dimension-agnostic in that it implements both `Primitive2d` and `Primitive3d`. This seems good on paper, but it can often be useful to have separate 2D and 3D versions of primitives. For example, one might want a two-dimensional capsule mesh. We can't really implement both 2D and 3D meshing for the same type using the upcoming `Meshable` trait (see bevyengine#11431). We also currently don't implement `Bounded2d` for `Capsule`, see bevyengine#11336 (comment). Having 2D and 3D separate at a type level is more explicit, and also more consistent with the existing primitives, as there are no other types that implement both `Primitive2d` and `Primitive3d` at the same time. ## Solution Rename `Capsule` to `Capsule3d` and add `Capsule2d`. `Capsule2d` implements `Bounded2d`. For now, I went for `Capsule2d` for the sake of consistency and clarity. Mathematically the more accurate term would be `Stadium` or `Pill` (see [Wikipedia](https://en.wikipedia.org/wiki/Stadium_(geometry))), but those might be less obvious to game devs. For reference, Godot has [`CapsuleShape2D`](https://docs.godotengine.org/en/stable/classes/class_capsuleshape2d.html). I can rename it if others think the geometrically correct name is better though. --- ## Changelog - Renamed `Capsule` to `Capsule3d` - Added `Capsule2d` with `Bounded2d` implemented --------- Co-authored-by: Alice Cecile <[email protected]>
Objective
Closes #10570.
#10946 added bounding volume types and traits, but didn't use them for anything yet. This PR implements
Bounded2dandBounded3dfor Bevy's primitive shapes.Solution
Implement
Bounded2dandBounded3dfor primitive shapes. This allows computing AABBs and bounding circles/spheres for them.For most shapes, there are several ways of implementing bounding volumes. I took inspiration from Parry's bounding volumes, Inigo Quilez, and figured out the rest myself using geometry. I tried to comment all slightly non-trivial or unclear math to make it understandable.
Parry uses support mapping (finding the farthest point in some direction for convex shapes) for some AABBs like cones, cylinders, and line segments. This involves several quat operations and normalizations, so I opted for the simpler and more efficient geometric approaches shown in Quilez's article.
Below you can see some of the bounding volumes working in 2D and 3D. Note that I can't conveniently add these examples yet because they use primitive shape meshing, which is still WIP.
2024-01-10.23-13-57.mp4
2024-01-13.01-26-14.mp4
Changelog
Bounded2d/Bounded3dfor primitive shapesfrom_point_cloudmethod for bounding volumes (used by many bounding implementations)point_cloud_2d/3d_centerandrotate_vec2utility functionsRegularPolygon::verticesmethod (used in regular polygon AABB construction)Triangle::circumcentermethod (used in triangle bounding circle construction)Extra
Do we want to implement
Bounded2dfor some "3D-ish" shapes too? For example, capsules are sort of dimension-agnostic and useful for 2D, so I think that would be good to implement. But a cylinder in 2D is just a rectangle, and a cone is a triangle, so they wouldn't make as much sense to me. A conical frustum would be an isosceles trapezoid, which could be useful, but I'm not sure if computing the 2D AABB of a 3D frustum makes semantic sense.