AABB Collision Detection: Why does the simple broadphase algorithm work better?

Question

Hi I am working on barrier detection.

I used this swept aabb algorithm http://www.gamedev.net/page/resources/_/technical/game-programming/swept-aabb-collision-detection-and-response-r3084.

It suggests two ways to detect collision. 1. broadphase: this is a shorter-algorithm which can only detect non-collision 100%. 2. In order to make sure that there is a collision, the more detailed swept aabb algorithm has to be executed right after the broadphase one.

Strangely the broadphase algorithm alone detects a barrier collision each and every time and the more precise one actually doesn't at all times.

What I do in my code is basically the following: My object moves towards the barrier with a constand speed vx = direction * speed * delta (that's the amount it moves each frame). So positionx = positionx + (direction * speed * delta) each frame. The broadphase-collision is calculated each frame by using vx and vy.

Are there some rare cases where the broadphase can actually detect a collision like in this one I described? Thanks so much for your help

Thanks so much

Answer 1

For reference, this is the broad-phasing:

The broad phase works 100% if the other object always overlaps with the swept area if it ever overlaps with the broad phase area. Usually this means axis-aligned barriers that are either infinite length, or whose ends/corners aren't inside the broad phase area.