Parallel rectilinear shortest paths with rectangular obstacles

Abstract

Let P be a simple rectilinear convex polygon of size O(n) inside which lie n pairwise disjoint rectangular rectilinear obstacles. We provide parallel techniques for computing rectilinear shortest paths that avoid the set of obstacles in P. Specifically, we compute descriptions of shortest paths in O(log2 n) time, with O(n2/log2n) processors in the CREWPRAM model if source and destination are on the boundary of P, with O(n2/log n) processors if the source is an obstacle vertex and the destination a vertex of P, and with O(n2) processors if both source and destination are obstacle vertices. The descriptions we compute enable one processor to obtain the path length of any query pair of vertices in constant time, or O(n/log n) processors to retrieve the shortest path itself in logarithmic time. If the two query points are arbitrary rather than vertices, then one processor takes O(log n) time (instead of constant time) for finding the path length. A number of other related shortest paths problems are solved. The techniques we use involve a fast computation of separator staircases and, most importantly, a scheme for partitioning the obstacles' boundaries into families in a way that ensures that the resulting path length matrices have a monotonicity property that is apparently absent before applying our partitioning scheme.