Sometimes Two Irrational Guards are Needed
Lucas Meijer, Tillmann Miltzow
TL;DR
This work proves that an optimal guarding of a simple monotone polygon with two guards may necessitate irrational coordinates, extending prior results that required three irrational guards and that one guard can be placed rationally. The authors construct a rational-coordinate polygon with two guard segments arranged to force a unique pair of irrational guard positions, and they prove that no two rational guards can guard the polygon. The approach leverages a detailed geometric construction with pockets and guard segments, together with an examination of the corresponding existential theory of the reals (ER/ETR) encoding to justify the necessity of irrational coordinates. The result sharpens our understanding of when irrational guard placements are unavoidable and provides a concrete benchmark for testing algorithms and discretization schemes in the art gallery problem.
Abstract
In the art gallery problem, we are given a closed polygon $P$, with rational coordinates and an integer $k$. We are asked whether it is possible to find a set (of guards) $G$ of size $k$ such that any point $p\in P$ is seen by a point in $G$. We say two points $p$, $q$ see each other if the line segment $pq$ is contained inside $P$. It was shown by Abrahamsen, Adamaszek, and Miltzow that there is a polygon that can be guarded with three guards, but requires four guards if the guards are required to have rational coordinates. In other words, an optimal solution of size three might need to be irrational. We show that an optimal solution of size two might need to be irrational. Note that it is well-known that any polygon that can be guarded with one guard has an optimal guard placement with rational coordinates. Hence, our work closes the gap on when irrational guards are possible to occur.
