Hi folks I am in the process of making a physics map abstraction from a generated terrain mesh. I am fairly new in the area and have come up with an initial step that I'd like to share and perhaps get some comments on. The idea is quite simple but it does impose some limitations with regards to flexibility. The limitation is based on the initial assumption that #1: The whole terrain is rectangular. #2 All vertices have a uniform distance along the respective planar axis (also meaning that all primitives have the same size when projected onto the 2D plane spanned by, say x and z axis). With this in mind I can now transform all tiles(quadrants into unit space and store the two primitives in a 2D array along with them. Along with then I compute the half plane for the diagonal separating the two primitives. From the primitives I compute the plane equation Ax+By+Cz+D=0. Also for sake of coarse pruning pretests I store a value indicating if the tile is traversable, not traversable or perhaps traversable (delegating the decision upwards). This is useful - for example - height computation which can now be done in 4 simple steps: - Transform the world coordinates into unit space (or into indices (i,j)). - Look up the Tile the array tiles[j]. - Find the primitive by half plane pruning. - Compute the height as y = (Ax0+Cz0+D)/B. I know that if I am to use this on a curving tunnel there will be some overhead but I find that acceptable. Also I will need to implement some multilayer segmentation but that is for another time. Using this abstraction I intend to look into navigation meshes and all that jazz. Personally I like this because it is simple to implement and I can basically use the technique on any mesh that conforms the two constraints previously mentioned. Is this a reasonable approach or does it have (serious) flaws? - Thanks