Collision between Sphere and Segment

one way is to take the projection of the vector from one vertex of the segment to the sphere center and project this on the segment. Then subtract the part along the segment from the vector to obtain the part perpendicular to the segment. If the length is bigger then the radius no collision.

Ya thank, but I already know that. Sorry if I difficulty to write well in english. :(

the problem is more what to do after! :P

If the test is true, I want to know the exact position tu push the sphere back.
I mean in collision, I have to consider this :

- I have the segment, wish I want to colide with.
- I have the radius of the sphere
- I have the last position of the sphere
- And the current position of the sphere

By calculating the distance with the two resulting segment I can know if there is a potential collision. AFTER this, I have to find where to put the sphere. Between the last position and the current. So it can fit perfectly with the segment.

Quote:

Ya thank, but I already know that. Sorry if I difficulty to write well in english. :( the problem is more what to do after! :P If the test is true, I want to know the exact position tu push the sphere back. I mean in collision, I have to consider this : - I have the segment, wish I want to colide with. - I have the radius of the sphere - I have the last position of the sphere - And the current position of the sphere By calculating the distance with the two resulting segment I can know if there is a potential collision. AFTER this, I have to find where to put the sphere. Between the last position and the current. So it can fit perfectly with the segment.

It sounds like you want a moving sphere vs. segment test, yes? This is a little tricky to implement, but isn't too difficult. First you need to intersect the sphere with the infinite line supporting the segment, which can be formulated similarly to a ray-cylinder test. If the sphere hits the line, move it to the time of intersection and see if the point where it touches the line is on the segment. If yes, you're done, else sweep against the appropriate endpoint, which can be formulated as a ray-sphere test.

If that sounds like more than you want to do, you can use a static test, assuming your sphere displacement is < its radius. You said you already know how to find the closest point on the segment to the sphere center, so:

Vector3 d = center - closest;
float length = d.Length();
d /= length;
center += (radius - length) * d;

That should resolve the collision. (Note that I didn't test for the degenerate case.)

If this is for sphere-triangle, there are some other algorithms you could use as well...

Heu no, I dont know the closest point! thats the problem.

But I juste resize the triangle with the radius of the sphere. This is not very accurate but well, doing the job

For closest point on a segment to a point, look here. (The algorithm is the same for 2d and 3d.)

Since its a triangle you could also do a quick test to see if sphere.position + sphere.radius > distance to the triangles plane. This would let you know if its even possible for the sphere to collide/intersect with the triangle or any of its segments at all.

1) find closest point on segment to sphere centre.
2) test if that point is inside the sphere.
3) jump with joy.

1) closest point on a edge to a point in space
1.a find closest point on the infinite line going through segment.
1.b constraint that point inside the segment boundaries.

1.a find closest point on the infinite line going through segment.
the closest point on a line to another point is the point that will be at the
prependicular on the segment to the point in space. simply put, you just project
that point on the line, like so (trivial linear algebra).

all the rest is simple.


for an algorithm about sphere-triangle collision, and in swept tests, I'll refer you to some of my collision demos.