Collision Problem

I don't have time to read your code, but it sounds like maybe you aren't transforming everything properly.

If you have a cube in model space (center 0,0,0) and you move that cube around in world space through the transform that you have stored in your class, then you need to be sure that transform is also being applied to the cube prior to the collision tests. There are a couple of ways of doing this. One is to lock the vertex buffer and multiple every vertex by the world transform, then set the world transform to Identity and don't use it anymore. The other way is to temporarily create a new cube, one that has every vertex transformed by the world matrix, and pass that cube into the collision routines.

To check collision between two cubes (let's say one is 10 units out in the x direction and the other is -5 units out), but both cubes are stored as (0,0,0) for their center (model coords), then you need to move each cube to its respective point in space, and then call the collision routines. What I mean is that you can't just pass in the untransformed model coordinates to the collision routines. You may have called the SetTransform(WORLD, cubeTransform) function, but that only moves the cube in the graphics pipeline, not in your cube class. I hope that makes sense.

Good luck,
Chris

thanks for your interest, i read your reply carefully. you say lock the vertex buffer of the mesh then multiply every vertex by the Matrix(World) here is my cube class. i derive two cubes from same class...

public class mesh_load
{
Mesh mesh = null;
Texture[] meshtextures = null;
Material[] meshmaterials = null;
public Vector3 min, max;

public Mesh Mesh_Load(Device device, string xfile)
{
ExtendedMaterial[] mtrl = null;
mesh = Mesh.FromFile(xfile, MeshFlags.Managed, device, out mtrl);
if(mtrl != null)
{
meshtextures = new Texture[mtrl.Length];
meshmaterials = new Material[mtrl.Length];
for(int i = 0; i < mtrl.Length; i++)
{
meshmaterials = mtrl.Material3D;
if(mtrl.TextureFilename != null)
{
meshtextures = TextureLoader.FromFile(device, mtrl.TextureFilename);
}
}
}
return mesh;
}

public void render(Device device, Vector3 position)
{
for(int i = 0; i < meshmaterials.Length; i++)
{
device.Material = meshmaterials;
device.SetTexture(0, meshtextures);
device.Transform.World = Matrix.Translation(position);//i set world transfrom to Matrix.Translation(position) because i move my cube. position is used as a different Vector3 in the other class. This different Vector3 takes value by keyboard entries.
mesh.DrawSubset(i);
}
collision(device);
}

void collision(Device device)
{
VertexBuffer vb = mesh.VertexBuffer;
try
{
GraphicsStream stm = vb.Lock(0, 0, LockFlags.None);
Geometry.ComputeBoundingBox(stm, mesh.NumberVertices, mesh.VertexFormat, out min, out max);
min.TransformCoordinate(device.Transform.World);
max.TransformCoordinate(device.Transform.World);
}
finally
{
vb.Unlock();
vb.Dispose();
}
}
}

Where and how should i multiply every vertex with world transform?
Thanks Murat...

u can use
Matrix.transfromcoordinate(vertices, worldtranformMatrix)

vertices shows the vertices of the bounding cube of the object
world transform represents the transformation u have applied on your object

the above api returns u the updated position of ur cube position as well

next perform your calculation to find the collision
hope it will work because i am using the same technique and its working

I think it's too slow to transform all the coordinates of an object and re-calculate the bounding sphere or the bounding box. If you are not doing any scaling, it's much faster to calculate a bounding sphere or a bounding box when the object is at (0,0,0) and then just transform the bounding sphere or the bounding box, not the object's vertices.

Think about it, a bounding sphere has just one vertex! (the center) so you only need to transform one vertex. A bounding box has more vertices and it probably needs to be rotated too, but it's still faster to transform a box than it is to transform hundreds of vertices.

In my game, I totally dumped the built-in DirectX Geometry.Boxboundprobe and Geometry.Sphereboundprobe methods and used my own calculatiuons, but I guess you can still use my method and the built-in DirectX functions.

if ur object is in motion which is normal case in all the games then transformations of all the vertices of ur mesh is the must case and in that case u must need to update the postions of ur bounding box or bounding sphere because the all the DirectX geometry related APIs return the result based on ur loaded mesh and not wrt to the transformed mesh position so must need to transform the coordinates of ur mesh as object is in motion

thanx all for your answers but i guess we misunderstood eachother...
i have two cubes, both derived from same class, in this class i make translation transformation to them via a Vector3 parameter in a mathod. now i have two cubes with different translations(model spaces) and i have a world space. i use Geometry.ComputeBoundingBox method for to get the min and max values for cubes in the same class. i want to collide these two cubes using Geometry.BoxBoundProbe method. i need to take both cubes in the same model space or in world space. i understood this. but how? any C# code for this? because when i tried to find some code i only found mouse picking examples, they're not same. i don't use mouse. i use just keyboard entries.

By the way for BoxBoundProbe method i need two vectors: rayposition and raydirection. my camera moves with the first cube, just behind it. i attached camera to it. i check whether this cube collides the other one.

please help...

Thanx

Murat

I solve the problem. Here's the solution: I said i have a class that my cubes are derived from. I added a Matrix struct to this class named "world". thus every object that is derived from this class had a Matrix struct named world. And i used this "world" in other class for the Vector3.TransfromCoordinate method. so my ray is transformed to my obstacle's world transform.

Thanx for everyone who replied my message.

Murat...