On the complexity of a single cell in certain arrangements of surfaces related to motion planning

We obtain near-quadratic upper bounds on the maximum combinatorial complexity of a single cell in certain arrangements ofn surfaces in 3-space where the lower bound for this quantity is Ω(n ²) or slightly larger. We prove a theorem that identifies a collection of topological and combinatorial conditions for a set of surface patches in space, which make the complexity of a single cell in an arrangement induced by these surface patches near-quadratic. We apply this result to arrangements related to motion-planning problems of two types of robot systems with three degrees of freedom and also to a special type of arrangements of triangles in space. The complexity of the entire arrangement in each case that we study can be Θ(n ³) in the worst case, and our single-cell bounds are of the formO(n ² α(n)), O(n ² logn), orO(n ² α(n)logn). The only previously known similar bounds are for the considerably simpler arrangements of planes or of spheres in space, where the bounds are Θ(n) and Θ(n ²), respectively. For some of the arrangements that we study we derive near-quadratic-time algorithms to compute a single cell. A preliminary version of this paper has appeared inProc. 7th ACM Symposium on Computational Geometry, North Conway, NH, 1991, pp. 314–323.