In this paper, we study the problem of efficiently reducing geometric shapes into other such shapes in a distributed setting through size-changing operations. We develop distributed algorithms using the reconfigurable circuit model to enable fast node-to-node communication. Our study considers two graph update models: the connectivity model and the adjacency model. Let $n$ denote the number of nodes and $k$ the number of turning points in the initial shape. In the connectivity model, we show that the system of nodes can reduce itself from any tree to a single node using only shrinking operations in $O(k \log n)$ rounds w.h.p. and any tree to its minimal (incompressible) form in $O(\log n)$ rounds with additional knowledge or $O(k \log n)$ without, w.h.p. We also give an algorithm to transform any tree to any topologically equivalent tree in $O(k \log n+\log^2 n)$ rounds w.h.p. if both shrinking and growth operations are available to the nodes. On the negative side, we show that one cannot hope for $o(\log^2 n)$-round transformations for all shapes of $O(\log n)$ turning points: for all reasonable values of $k$, there exists a pair of geometrically equivalent paths of $k$ turning points each, such that $\Omega(k\log n)$ rounds are required to reduce one to the other. In the adjacency model, we show that the system can reduce itself from any connected shape to a single node using only shrinking in $O(\log n)$ rounds w.h.p.
翻译:暂无翻译