Original Post
Hey,
I have two possible implementations for the collision detection/physics in my game with static terrain, and I'm trying to figure out which one you guys might recommend more.
1. Realistic physics
When the character walks into a wall or falls onto the ground, they actually penetrate the ground, and a force is applied to the character to push him out, just as the real world works. The farther the ground is displaced (not visually), the more the force onto the character which pushes them out of the ground. How would you calculate how deep the character has penetrated, and which direction to apply the outward force?
2. Estimate boundaries
When the character walks into a wall or falls onto the ground, the game checks the bounding box of the character versus all nearby platforms and the heightmap. If it sees that a character has penetrated something, it uses the (assumed to be acceptable) previous coordinates of the character and the coordinates the character would like to travel to, and calculates the precise point in between that is as close to the wall/platform as possible without actually colliding with it.
-- FYI
I'm currently using a heightmap for the absolute lowest position of the terrain, and arbitrary meshes for platforms/structures on top of the terrain. So far, I've implemented #2 with a heightmap, assuming the character is an infinitely small point (no platforms yet). I've also heard some stuff about a number of spheres working better than a bounding box for collision detection. Ultimately I'd be able to implement #1 with spheres for a character (rather than a bounding box), as I've heard these are the best simple solutions for a 3D indie game.
Another small question: if the character is on the slope of a heightmap, I'd like to orient him in a way such that he stands straight up in the direction of the normal vector from the slope he stands on. However, what if he's standing on multiple triangles simultaneously (characters aren't infinitely small, you don't want him to jump from one normal to the next, it should be a transition), should I take a weighted average of the normals of all triangles he's standing over?
Ideas/comments? Thanks!
I have two possible implementations for the collision detection/physics in my game with static terrain, and I'm trying to figure out which one you guys might recommend more.
1. Realistic physics
When the character walks into a wall or falls onto the ground, they actually penetrate the ground, and a force is applied to the character to push him out, just as the real world works. The farther the ground is displaced (not visually), the more the force onto the character which pushes them out of the ground. How would you calculate how deep the character has penetrated, and which direction to apply the outward force?
2. Estimate boundaries
When the character walks into a wall or falls onto the ground, the game checks the bounding box of the character versus all nearby platforms and the heightmap. If it sees that a character has penetrated something, it uses the (assumed to be acceptable) previous coordinates of the character and the coordinates the character would like to travel to, and calculates the precise point in between that is as close to the wall/platform as possible without actually colliding with it.
-- FYI
I'm currently using a heightmap for the absolute lowest position of the terrain, and arbitrary meshes for platforms/structures on top of the terrain. So far, I've implemented #2 with a heightmap, assuming the character is an infinitely small point (no platforms yet). I've also heard some stuff about a number of spheres working better than a bounding box for collision detection. Ultimately I'd be able to implement #1 with spheres for a character (rather than a bounding box), as I've heard these are the best simple solutions for a 3D indie game.
Another small question: if the character is on the slope of a heightmap, I'd like to orient him in a way such that he stands straight up in the direction of the normal vector from the slope he stands on. However, what if he's standing on multiple triangles simultaneously (characters aren't infinitely small, you don't want him to jump from one normal to the next, it should be a transition), should I take a weighted average of the normals of all triangles he's standing over?
Ideas/comments? Thanks!
