On the Shortest Separating Cycle

Abstract

According to a result of Arkin~\etal~(2016), given $n$ point pairs in the plane, there exists a simple polygonal cycle that separates the two points in each pair to different sides; moreover, a $O(\sqrt{n})$-factor approximation with respect to the minimum length can be computed in polynomial time. Here the following results are obtained: (I)~We extend the problem to geometric hypergraphs and obtain the following characterization of feasibility. Given a geometric hypergraph on points in the plane with hyperedges of size at least $2$, there exists a simple polygonal cycle that separates each hyperedge if and only if the hypergraph is $2$-colorable. (II)~We extend the $O(\sqrt{n})$-factor approximation in the length measure as follows: Given a geometric graph $G=(V,E)$, a separating cycle (if it exists) can be computed in $O(m+ n\log{n})$ time, where $|V|=n$, $|E|=m$. Moreover, a $O(\sqrt{n})$-approximation of the shortest separating cycle can be found in polynomial time. Given a geometric graph $G=(V,E)$ in $\mathbb{R}^3$, a separating polyhedron (if it exists) can be found in $O(m+ n\log{n})$ time, where $|V|=n$, $|E|=m$. Moreover, a $O(n^{2/3})$-approximation of a separating polyhedron of minimum perimeter can be found in polynomial time. (III)~Given a set of $n$ point pairs in convex position in the plane, we show that a $(1+\varepsilon)$-approximation of a shortest separating cycle can be computed in time $n^{O(\varepsilon^{-1/2})}$. In this regard, we prove a lemma on convex polygon approximation that is of independent interest.