[Paper Review] Smoothed Analysis of the Art Gallery Problem

Michael Gene Dobbins, Andreas F. Holmsen+1 more|arXiv (Cornell University)|Nov 3, 2018

Computational Geometry and Mesh Generation50 references5 citations

TL;DR

This paper applies smoothed analysis to the Art Gallery Problem, demonstrating that under small random perturbations of polygon vertices, optimal guard positions typically require only a small number of bits to describe—implying rational coordinates suffice in practice. The key result is that the Naive Algorithm (guessing guard positions and verifying visibility) runs in expected non-deterministic polynomial time, suggesting algebraic methods are rarely needed for typical instances, despite the problem's ∃ℝ-completeness.

ABSTRACT

In the Art Gallery Problem we are given a polygon $P\subset [0,L]^2$ on $n$ vertices and a number $k$. We want to find a guard set $G$ of size $k$, such that each point in $P$ is seen by a guard in $G$. Formally, a guard $g$ sees a point $p \in P$ if the line segment $pg$ is fully contained inside the polygon $P$. The history and practical findings indicate that irrational coordinates are a "very rare" phenomenon. We give a theoretical explanation. Next to worst case analysis, Smoothed Analysis gained popularity to explain the practical performance of algorithms, even if they perform badly in the worst case. The idea is to study the expected performance on small perturbations of the worst input. The performance is measured in terms of the magnitude $δ$ of the perturbation and the input size. We consider four different models of perturbation. We show that the expected number of bits to describe optimal guard positions per guard is logarithmic in the input and the magnitude of the perturbation. This shows from a theoretical perspective that rational guards with small bit-complexity are typical. Note that describing the guard position is the bottleneck to show NP-membership. The significance of our results is that algebraic methods are not needed to solve the Art Gallery Problem in typical instances. This is the first time an $\exists\mathbb{R}$-complete problem was analyzed by Smoothed Analysis.

Motivation & Objective

• To explain why irrational coordinates are exceptionally rare in optimal Art Gallery Problem solutions.
• To analyze the practical feasibility of exact algorithms for the Art Gallery Problem, given its ∃ℝ-completeness.
• To show that small perturbations of polygon inputs typically yield solutions with rational, low-complexity guard positions.
• To provide a theoretical justification for the empirical success of non-algebraic algorithms in solving medium-sized Art Gallery instances.
• To explore whether smoothed analysis can overcome the inherent hardness of ∃ℝ-complete problems in computational geometry.

Proposed method

• Four perturbation models are analyzed: Minkowski-Inflation, Edge-Inflation, Edge-Perturbation, and Vertex-Perturbation, to model small random changes to polygon vertices.
• The expected bit complexity of describing optimal guard positions is bounded as O(log(nL/δ)) for three models and O(log(nL/(δβ))) for Vertex-Perturbation, where β is the minimum interior or exterior angle.
• A discrete Edge-Inflation model is introduced to show that instances can be solved in non-deterministic polynomial time with high probability.
• The Naive Algorithm—guessing guard positions and verifying visibility—is analyzed under smoothed conditions, showing it runs in expected non-deterministic polynomial time.
• For continuous perturbations, the algorithm requires O(n⁴) additional time on a real RAM to verify guard correctness due to real-number coordinates.
• Theoretical results are extended to show correctness of approximation algorithms in the smoothed analysis setting.

Experimental results

Research questions

• RQ1 Does smoothed analysis explain why irrational coordinates are so rare in optimal Art Gallery solutions?
• RQ2 Can the Naive Algorithm be expected to solve typical Art Gallery instances in non-deterministic polynomial time under small perturbations?
• RQ3 Is there a polynomial-sized candidate set of guard positions that can be used in smoothed analysis for the Art Gallery Problem?
• RQ4 Why do practical implementations spend most time on geometric routines rather than the NP-hard set-cover part?
• RQ5 Can smoothed analysis be generalized to other ∃ℝ-complete problems like motion planning or Nash equilibrium computation?

Key findings

• Under small random perturbations, the expected number of bits needed to describe optimal guard positions is O(log(nL/δ)) for three perturbation models, indicating rational guards with small bit complexity are typical.
• For Vertex-Perturbation, the bit complexity is O(log(nL/(δβ))), where β is the minimum interior or exterior angle, showing angle structure affects solution complexity.
• The Naive Algorithm runs in expected non-deterministic polynomial time under discrete Edge-Inflation, implying algebraic methods are not required for typical instances.
• In the continuous perturbation model, the Naive Algorithm requires O(n⁴) additional time on a real RAM to verify guard correctness, due to real-number coordinates.
• The results imply that algebraic methods are not essential for solving the Art Gallery Problem in typical cases, despite its ∃ℝ-completeness.
• The study establishes the first smoothed analysis of an ∃ℝ-complete problem, showing that such problems can be tractable in practice under mild randomness.