Machine-learning based collision detection for objects in virtual environments

1. A method comprising:

processing, using a machine learning model, one or more parameters associated with a ray cast in virtual environment comprising an object, the machine learning model configured to generate an approximated mesh representing the object based on geometric surface data associated with the object;

generating a proximity score for potential intersections, using the machine learning model, by calculating distances between the ray and relevant portions of the approximated mesh, wherein a collision is indicated when the proximity score satisfies a threshold indicating an intersection with the approximated mesh;

responsive to a collision between the ray and the object as indicated by the proximity score, generating, using the machine learning model, collision data associated with the ray and the object; and

updating, based on the collision data, at least one of the object or the virtual environment to indicate the collision occurred.

2. The method of claim 1 , further comprising:

selecting the machine learning model from a plurality of machine learning models based on the object, wherein different machine learning models are associated with different objects.

3. The method of claim 1 , wherein updating the at least one of the object or the virtual environment comprises:

rendering the object in the virtual environment based on the collision data generated using the machine learning model.

4. The method of claim 1 , further comprising:

obtaining the one or more parameters associated with the ray as input to the machine learning model in response to:

determining the ray intersects a bounding volume of the object based on an origin point of the ray, a direction vector of the ray, and extent values of the bounding volume.

5. The method of claim 4 , further comprising:

determining a point on the bounding volume intersected by the ray;

transforming the point from world-space coordinates to local-space coordinates of the bounding volume to generate a local-space intersection point;

normalizing the local-space intersection point to lay on a surface of a unit volume to generate a normalized origin point for the ray on the bounding volume of the object; and

selecting the normalized origin point as one of the one or more parameters of the ray.

6. The method of claim 5 , further comprises:

normalizing the direction vector of the ray to the point on the bounding volume intersected by the ray;

transforming the direction vector from world-space coordinates to local-space coordinates of the bounding volume to generate a normalized local-space direction vector for the ray; and

selecting the normalized local-space direction vector as one of the one or more parameters of the ray.

7. The method of claim 1 , wherein the machine learning model is a trained based on casting rays from a point on or within a bounding volume of the object.

8. The method of claim 1 , wherein the collision data comprises:

an indication whether the ray collides with the object, and

if a collision occurred between the ray and the object, the collision data further comprises information associated with collision, the information comprising at least one of:

a location on the object where the ray intersects;

a normal vector associated with the collision;

color coordinates of the location on the object where the ray intersects; or

texture coordinates of the location on the object where the ray intersects.

9. An apparatus comprising:

a processor; and

a collision detection unit operatively coupled to the processor, the collision detection unit being configured to:

process, using a machine learning model, one or more parameters associated with a ray cast in virtual environment comprising an object, the machine learning model configured to generate an approximated mesh representing the object based on geometric surface data associated with the object;

generate, using the machine learning model, a proximity score for potential intersections by calculating distances between the ray and relevant portions of the approximated mesh, wherein a collision is indicated when the proximity score satisfies a threshold indicating an intersection with the approximated mesh; and

responsive to a collision between the ray and the object, generate, using the machine learning model, collision data associated with the ray and the object,

the processor configured to update, based on the collision data, at least one of the object or the virtual environment to indicate the collision occurred.

10. The apparatus of claim 9 , wherein the collision detection unit is configured to:

select the machine learning model from a plurality of machine learning models based on the object, wherein different machine learning models are associated with different objects.

11. The apparatus of claim 9 , further comprising:

a collision resolution unit configured to resolve a collision between the ray and the object, by rendering the object in the virtual environment based on the collision data generated using the machine learning model.

12. The apparatus of claim 11 , wherein the collision detection unit is configured to:

obtain the one or more parameters associated with the ray as input to the machine learning model in response to:

determining the ray intersects a bounding volume of the object based on an origin point of the ray, a direction vector of the ray, and extent values of the bounding volume.

13. The apparatus of claim 12 , wherein collision detection unit is further configured to:

determine a point on the bounding volume intersected by the ray;

transform the point from world-space coordinates to local-space coordinates of the bounding volume to generate a local-space intersection point;

normalize the local-space intersection point to lay on a surface of a unit volume to generate a normalized origin point for the ray on the bounding volume of the object; and

select the normalized origin point as one of the one or more parameters of the ray.

14. The apparatus of claim 13 , wherein collision detection unit is further configured to:

normalize the direction vector of the ray to the point on the bounding volume intersected by the ray;

transform the direction vector from world-space coordinates to local-space coordinates of the bounding volume to generate a normalized local-space direction vector for the ray; and

select the normalized local-space direction vector as one of the one or more parameters of the ray.

15. The apparatus of claim 9 , wherein the machine learning model is a trained based on casting rays from a point on or within a bounding volume of the object.

16. The apparatus of claim 9 , wherein the collision data comprises:

an indication whether the ray collides with the object, and

if a collision occurred between the ray and the object, the collision data further comprises information associated with collision, the information comprising at least one of:

a location on the object where the ray intersects;

a normal vector associated with the collision;

color coordinates of the location on the object where the ray intersects; or

texture coordinates of the location on the object where the ray intersects.

17. A method comprising:

casting a ray within a virtual environment comprising an object;

determining a local-space intersection point of the ray on a bounding volume of the object;

setting the local-space intersection point as an origin point of the ray;

determining a direction vector of the ray based on the local-space intersection point;

providing the origin point and the direction vector of the ray to a machine learning model trained to generate an approximated mesh for the object based on geometric surface data associated with the object;

processing, using the machine learning model, the origin point and the direction vector by:

excluding regions of the approximated mesh where collisions are unlikely based on a direction and a position of the ray;

calculating distances between the ray and relevant portions of the approximated mesh; and

responsive to the calculated distances indicating that the ray collides with the object, generating, using the machine learning model, a collision indication; and

updating, based on the collision indication, at least one of the object or the virtual environment to indicate the collision occurred.

18. The method of claim 17 , wherein generating the collision indication comprises:

generating, by the machine learning model, collision data associated with the ray and the object in response to the calculated distances indicating that the ray collides with the object.

19. The method of claim 17 , wherein setting the local-space intersection point as an origin point of the ray comprises:

normalizing the local-space intersection point to lay on a surface of a unit volume.

20. The method of claim 17 , wherein determining the direction vector of the ray based on the local-space intersection point comprises:

normalizing the direction vector of the ray to the local-space intersection point; and

transforming the normalized direction vector from world-space coordinates to local-space coordinates of the bounding volume.