NeHe OpenGL Lesson30 – Collision Detection

This sample shows us how to work with collision detection and response. At first, it provide some ideas about how to determine whether a intersection happened between a ray and a plane, a ray and a cylinder, sphere and sphere. Then set up the collision response, calculate reflection direction and speed based on the collision point normal.
At last, a very practical method provided here that simulating the whole moving stuffs always with the nearest collision time. Generally, we would some update function and with a double type delta time as input parameter.  And we want the object’s position information after the delta time elapsed with this function. We may predict whether some collision happen or not during such delta time, then calculate the collision response basing on the prediction, and advance to the next frame.  But the problem is that some sequential several collisions may happen during the frame time. A better solution to handle this is that we always try to find the nearest the collision point, then moving all objects to this collision point, then advance the collision time until the end of the frame time. Before we start calculate the new positions for objects, we need to update it’s velocities first. Here are the pseudo code that does what you really need:

void CollisionHandle(float deltaTime)
{
    compute moving object velocity;
 
    totalTime = 0.0f;
    while (totalTime < deltaTime)
    {
        CollisionTime = {compute nearest collision time};
        if ( CollisionTime + totalTime > deltaTime )
        {
            MovingTime = (deltaTime - totalTime );
            { Advance all moving objects with "MovingTime" };
            break;
        }
        else
        {
            { Advance all non-nearest moving objects with "CollisionTime" };
            { Collision response calculation happen here };
            totalTime += CollisionTime;
        }
    }
}

Well, this method may works well at some situation. But you may already notice that this method will make the whole moving & collision process become sequential. Actually, some of them will not collision at all during the frame time.  Maybe we could separate moving objects into different groups, and make them run parallel. Another question is what we should do when the shape of the object become very complicated, may be more than one collision point found at the same time. Well, why do not we ask help for some professional physics library instead of writing our own library.

 

The explosion effect that used here were not billboards, that are two quads(with front and back surfaces) which perpendicular with each other. With this method, we do not need to get the camera orientation at all. This is another way to simulate some trees.

 

The full source code could be found here.