I'm pretty unfamiliar with data structures used in scenegraphs for 3D games. Basically I'm interested in nothing too heavy duty, just semi-smart culling of objects. Our player will be flying around in a space ship--and space is relatively sparse. In one "sector," there might be a planet off in the distance, a looming rotating space station, asteroid field off in the distance, and various other ships milling about. I'm looking for the most intelligent way to divide up all that space, for both rendering and collision detection. Most importantly, if the player isn't facing towards the space station, don't draw it. I'm thinking the easiest method would just be to define a bounding sphere with each "Renderable" object. But does anyone have any experience in this area? What would the base class for a Renderable look like? Should everything be a tree? Should model data and bounding information be tightly coupled?