On Stable Approximation Algorithms for Geometric Coverage Problems

TL;DR

This paper develops a more stable approximation algorithm for the dynamic Max Cover by Unit Disks problem, achieving near-optimal coverage with fewer changes, and explores the limitations for non-fat objects like segments.

Contribution

It introduces a $k_ ext{varepsilon}$-stable $(1- ext{varepsilon})$-approximation algorithm for dynamic geometric coverage, improving stability bounds and extending to other fat objects.

Findings

Achieves $k_ ext{varepsilon}$-stability with $O(1/ ext{varepsilon}^3)$ changes.

Extends results to arbitrarily-oriented fat shapes like squares and ellipses.

Shows limitations for non-fat objects, requiring many changes for approximation.

Abstract

Let $P$ be a set of points in the plane and let $m$ be an integer. The goal of Max Cover by Unit Disks problem is to place $m$ unit disks whose union covers the maximum number of points from~$P$. We are interested in the dynamic version of Max Cover by Unit Disks problem, where the points in $P$ appear and disappear over time, and the algorithm must maintain a set \cDalg of $m$ disks whose union covers many points. A dynamic algorithm for this problem is a $k$-stable $\alpha$-approximation algorithm when it makes at most $k$ changes to \cDalg upon each update to the set $P$ and the number of covered points at time $t$ is always at least $\alpha \cdot \opt(t)$, where $\opt(t)$ is the maximum number of points that can be covered by m disks at time $t$. We show that for any constant $\varepsilon>0$, there is a $k_{\varepsilon}$-stable $(1-\varepsilon)$-approximation algorithm for the dynamic Max Cover by Unit Disks problem, where $k_{\varepsilon}=O(1/\varepsilon^3)$. This improves the stability of $\Theta(1/\eps^4)$ that can be obtained by combining results of Chaplick, De, Ravsky, and Spoerhase (ESA 2018) and De~Berg, Sadhukhan, and Spieksma (APPROX 2023). Our result extends to other fat similarly-sized objects used in the covering, such as arbitrarily-oriented unit squares, or arbitrarily-oriented fat ellipses of fixed diameter. We complement the above result by showing that the restriction to fat objects is necessary to obtain a SAS. To this end, we study the Max Cover by Unit Segments problem, where the goal is to place $m$ unit-length segments whose union covers the maximum number of points from $P$. We show that there is a constant $\varepsilon^* > 0$ such that any $k$-stable $(1 + \varepsilon^*)$-approximation algorithm must have $k=\Omega(m)$, even when the point set never has more than four collinear points.