Collision in 3d car game
Author
I am making a 3d car driving game.The car physics are almost completed and
almost realistic.Now I need to incorporate car-car collision effects.I have
seen that GTA(Grand Theft Auto) and some other games use a collision model
that is made up of only spheres.I know how to detect collisions with a collision model like this but WHAT I DONT KNOW IS HOW TO CALCULATE THE COLLISION NORMALS
In these models the actual collision normal may not always be the normal corresponding to the spheres that collided. I would also like to know which is
the best method to implement a realistics car-car collision in a 3d driving game
I'm not a specialist in collision effects but I think that this project will be very interesting in your case : http://rars.sourceforge.net/
I hope this help
Yannick
I hope this help
Yannick
if you alreay have spheres, then it's sorted.
In Rollcage, for example, cars are made of 6 spheres (plus 1 per wheel), and the collision normals are nothing more than when doing sphere/sphere collisions.
That's a decent approximation, if rather low tech (Rollcage was made for the PSX and Pentium 1 systems). You can always increase the sphere density to get a better approximation. YOu can also build a sphere tree to speed up the test and increase the density.
Else, you can use a collection of boxes or convex hulls, and perform collisions with those. Obviously, you'll need a much more capable collision system (Swept GJK, or advanced SAT).
I'd stick with spheres for a while. It should take about a week to experiement and get decent results. If you're not happy, look into convex hull collisions. But that would take at least a month to get right.
Spheres also have the advantage of not having sharp angles, which makes collisions more forgiving and less harsh.
In Rollcage, for example, cars are made of 6 spheres (plus 1 per wheel), and the collision normals are nothing more than when doing sphere/sphere collisions.
That's a decent approximation, if rather low tech (Rollcage was made for the PSX and Pentium 1 systems). You can always increase the sphere density to get a better approximation. YOu can also build a sphere tree to speed up the test and increase the density.
Else, you can use a collection of boxes or convex hulls, and perform collisions with those. Obviously, you'll need a much more capable collision system (Swept GJK, or advanced SAT).
I'd stick with spheres for a while. It should take about a week to experiement and get decent results. If you're not happy, look into convex hull collisions. But that would take at least a month to get right.
Spheres also have the advantage of not having sharp angles, which makes collisions more forgiving and less harsh.
Author
Thank you
I have decided to proceed with the sense sphere method.
Can you please brief the algorithm for that
I have decided to proceed with the sense sphere method.
Can you please brief the algorithm for that
there are plenty of samples for sphere-sphere collisions, and triangle-sphere collisions.
Sphere-sphere is relatively trivial.
here is one.
http://www.gamedev.net/reference/articles/article1234.asp
triangle-sphere is a little bit trickier. That would be for car-mesh collisions (like buildings, and so forth).
so here is another one.
http://www.fluidstudios.com/?freeware
http://www.fluidstudios.com/pub/FluidStudios/CollisionDetection/Fluid_Studios_Generic_Collision_Detection_for_Games_Using_Ellipsoids.pdf
If you read the docs, the actual procedure is relatively simple. Just a little but of linear algebra.
You can apply the principles of the second paper for the sphere-pshere colision response.
Sphere-sphere is relatively trivial.
here is one.
http://www.gamedev.net/reference/articles/article1234.asp
triangle-sphere is a little bit trickier. That would be for car-mesh collisions (like buildings, and so forth).
so here is another one.
http://www.fluidstudios.com/?freeware
http://www.fluidstudios.com/pub/FluidStudios/CollisionDetection/Fluid_Studios_Generic_Collision_Detection_for_Games_Using_Ellipsoids.pdf
If you read the docs, the actual procedure is relatively simple. Just a little but of linear algebra.
You can apply the principles of the second paper for the sphere-pshere colision response.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
