okay here's just my proposed plan for collision detection, i just want to get some feedback on it (possible problems, good things, bad things, slow things, etc.) my collision detection system works by creating axis aligned bounding boxes around all entities. then during each frame, i update the boxes with physics to their new wanted position. to account for possible 'jumping through objects' (when one object moves so quickly that its new position is completly on the other side of an object without being detected that it went through), i will build a new model based off the two bounding boxes that encompasses the whole area that the model might have passed through. so in 2d it would look something like this: before:

-----
|old|
|   |
-----
    -----
    |new|
    |   |
    -----
  

after:

-----
\     \    \ 
  \       -----
  

using this new model i can easily detect those jump through collisions, not only that but i can backtrack to exactly what time/where they collided and go from there. whenever there is a collision however, i transform the actual high detail model to where the collision took place to do a full polygon to polygon collision detection. this is what i haven't figured out yet... with the aabb model i built from the delta position, it would give me a range of where the actual models could have intersected, but i do not know how i would test a range with the actual models. any ideas? [edited by - Rocket05 on March 16, 2003 10:35:11 AM] [edited by - Rocket05 on March 16, 2003 10:36:04 AM] okay apparently i cant get the second drawing to show up right in the forums, but basically think of a 6 sided shape, made up of the upper left two sides of the old position, and the bottom right sides of the new position, and the two other sides made up of lines made between the old & new. [edited by - Rocket05 on March 16, 2003 10:38:20 AM]