Up to date
This page is up to date for Godot
If you still find outdated information, please open an issue.
The types of collision shapes available in 2D in Godot.
Using an image converted to a polygon as a collision shape.
Performance considerations regarding 2D collisions.
You can define the shape of a PhysicsBody2D by adding one or more CollisionShape2Ds or CollisionPolygon2Ds as child nodes. Note that you must add a Shape2D resource to collision shape nodes in the Inspector dock.
When you add multiple collision shapes to a single PhysicsBody2D, you don't have to worry about them overlapping. They won't "collide" with each other.
SeparationRayShape2D (designed for characters)
WorldBoundaryShape2D (infinite plane)
你可以使用一个或多个原始形状来表示大多数较小物体的碰撞. 然而, 对于更复杂的物体, 如大型船舶或整个水平面, 你可能需要凸形或凹形来代替. 下面会有更多的介绍.
We recommend favoring primitive shapes for dynamic objects such as RigidBodies and CharacterBodies as their behavior is the most reliable. They often provide better performance as well.
Godot currently doesn't offer a built-in way to create 2D convex collision shapes. This section is mainly here for reference purposes.
Convex collision shapes are a compromise between primitive collision shapes and concave collision shapes. They can represent shapes of any complexity, but with an important caveat. As their name implies, an individual shape can only represent a convex shape. For instance, a pyramid is convex, but a hollow box is concave. To define a concave object with a single collision shape, you need to use a concave collision shape.
根据对象的复杂程度, 可能要通过使用多个凸形而不是一个凹形碰撞形状来获得更好的性能.Godot可以使用 凸分解 来生成与空心物体大致匹配的凸形. 请注意, 在一定数量的凸形之后, 就没有了这种性能优势, 对于大而复杂的对象, 如整个关卡, 建议使用凹形代替.
Concave collision shapes, also called trimesh collision shapes, can take any form, from a few triangles to thousands of triangles. Concave shapes are the slowest option but are also the most accurate in Godot. You can only use concave shapes within StaticBodies. They will not work with CharacterBodies or RigidBodies unless the RigidBody's mode is Static.
即使凹形提供了最准确的 碰撞, 但触碰信息的精度可能不如基础形状.
When not using TileMaps for level design, concave shapes are the best approach for a level's collision.
You can configure the CollisionPolygon2D node's build mode in the inspector. If it is set to Solids (the default), collisions will include the polygon and its contained area. If it is set to Segments, collisions will only include the polygon edges.
You can generate a concave collision shape from the editor by selecting a Sprite2D and using the Sprite2D menu at the top of the 2D viewport. The Sprite2D menu dropdown exposes an option called Create CollisionPolygon2D Sibling. Once you click it, it displays a menu with 3 settings:
Simplification: Higher values will result in a less detailed shape, which improves performance at the cost of accuracy.
Shrink (Pixels): Higher values will shrink the generated collision polygon relative to the sprite's edges.
Grow (Pixels): Higher values will grow the generated collision polygon relative to the sprite's edges. Note that setting Grow and Shrink to equal values may yield different results than leaving both of them on 0.
If you have an image with many small details, it's recommended to create a simplified version and use it to generate the collision polygon. This can result in better performance and game feel, since the player won't be blocked by small, decorative details.
To use a separate image for collision polygon generation, create another Sprite2D, generate a collision polygon sibling from it then remove the Sprite2D node. This way, you can exclude small details from the generated collision.
You aren't limited to a single collision shape per PhysicsBody. Still, we recommend keeping the number of shapes as low as possible to improve performance, especially for dynamic objects like RigidBodies and CharacterBodies. On top of that, avoid translating, rotating, or scaling CollisionShapes to benefit from the physics engine's internal optimizations.
When using a single non-transformed collision shape in a StaticBody, the engine's broad phase algorithm can discard inactive PhysicsBodies. The narrow phase will then only have to take into account the active bodies' shapes. If a StaticBody has many collision shapes, the broad phase will fail. The narrow phase, which is slower, must then perform a collision check against each shape.