Approximating the shortest path problem with scenarios

Adam Kasperski,Pawel Zielinski

2024-09-17

Abstract:This paper discusses the shortest path problem in a general directed graph with $n$ nodes and $K$ cost scenarios (objectives). In order to choose a solution, the min-max criterion is applied. The min-max version of the problem is hard to approximate within $\Omega(\log^{1-\epsilon} K)$ for any $\epsilon>0$ unless NP$\subseteq \text{DTIME}(n^{\text{polylog} \,n})$ even for arc series-parallel graphs and within $\Omega(\log n/\log\log n)$ unless NP$\subseteq \text{ZPTIME}(n^{\log\log n})$ for acyclic graphs. The best approximation algorithm for the min-max shortest path problem in general graphs, known to date, has an approximation ratio of~$K$. In this paper, an $\widetilde{O}(\sqrt{n})$ flow LP-based approximation algorithm for min-max shortest path in general graphs is constructed. It is also shown that the approximation ratio obtained is close to an integrality gap of the corresponding flow LP relaxation.

Data Structures and Algorithms

What problem does this paper attempt to address?

This paper aims to solve a classic problem in graph theory - the shortest path problem, but it is discussed in the multi - scenario cost setting in an uncertain environment. Specifically, the paper focuses on how to select a shortest path in a general directed graph with \(n\) nodes and \(K\) cost scenarios to minimize the path with the maximum cost in all possible scenarios. This is known as the min - max shortest path problem.

The paper points out that for arc series - parallel graphs, even if \(K\) is a constant, this problem is also difficult to be approximately solved within \(\Omega(\log^{1 - \epsilon}K)\), unless \(NP\subseteq DTIME(n^{\text{polylog }n})\); and for acyclic graphs, it is also difficult to be approximately solved within \(\Omega(\log n / \log \log n)\), unless \(NP\subseteq ZPTIME(n^{\log \log n})\).

Currently, the known best approximation algorithm has an approximation ratio of \(K\) in general graphs. This paper proposes a \(\tilde{O}(\sqrt{n})\) approximation algorithm based on flow linear programming (flow LP) to solve the min - max shortest path problem in general graphs, and proves that the obtained approximation ratio is close to the integrality gap of the corresponding flow linear programming relaxation.

In short, this paper attempts to improve existing solutions by proposing a new approximation algorithm to more effectively handle the shortest path problem in an uncertain environment.