Efficient Exact Algorithms for Minimum Covering of Orthogonal Polygons with Squares

Abstract

The Orthogonal Polygon Covering with Squares (OPCS) problem takes as input an orthogonal polygon $P$ without holes with $n$ vertices, where vertices have integral coordinates. The aim is to find a minimum number of axis-parallel, possibly overlapping squares which lie completely inside $P$, such that their union covers the entire region inside $P$. Aupperle et. al~\cite{aupperle1988covering} provide an $\mathcal O(N^{1.5})$-time algorithm to solve OPCS for orthogonal polygons without holes, where $N$ is the number of integral lattice points lying in the interior or on the boundary of $P$. Designing algorithms for OPCS with a running time polynomial in $n$ (the number of vertices of $P$) was discussed as an open question in \cite{aupperle1988covering}, since $N$ can be exponentially larger than $n$. In this paper we design a polynomial-time exact algorithm for OPCS with a running time of $\mathcal O(n^{14})$. We also consider the following structural parameterized version of the problem. A knob in an orthogonal polygon is a polygon edge whose both endpoints are convex polygon vertices. Given an input orthogonal polygon with $n$ vertices and $k$ knobs, we design an algorithm for OPCS with running time $\mathcal O(n^2 + k^{14} \cdot n)$. In \cite{aupperle1988covering}, the Orthogonal Polygon with Holes Covering with Squares (OPCSH) problem is also studied where orthogonal polygon could have holes, and the objective is to find a minimum square covering of the input polygon. This is shown to be NP-complete. We think there is an error in the existing proof in \cite{aupperle1988covering}, where a reduction from Planar 3-CNF is shown. We fix this error in the proof with an alternate construction of one of the gadgets used in the reduction, hence completing the proof of NP-completeness of OPCSH.