In the Disjoint Paths problem, the input is an undirected graph $G$ on $n$ vertices and a set of $k$ vertex pairs, $\{s_i,t_i\}_{i=1}^k$, and the task is to find $k$ pairwise vertex-disjoint paths connecting $s_i$ to $t_i$. The problem was shown to have an $f(k)n^3$ algorithm by Robertson and Seymour. In modern terminology, this means that Disjoint Paths is fixed parameter tractable (FPT), parameterized by the number of vertex pairs. This algorithm is the cornerstone of the entire graph minor theory, and a vital ingredient in the $g(k)n^3$ algorithm for Minor Testing (given two undirected graphs, $G$ and $H$ on $n$ and $k$ vertices, respectively, the objective is to check whether $G$ contains $H$ as a minor). All we know about $f$ and $g$ is that these are computable functions. Thus, a challenging open problem in graph algorithms is to devise an algorithm for Disjoint Paths where $f$ is single exponential. That is, $f$ is of the form $2^{{\sf poly}(k)}$. The algorithm of Robertson and Seymour relies on topology and essentially reduces the problem to surface-embedded graphs. Thus, the first major obstacle that has to be overcome in order to get an algorithm with a single exponential running time for Disjoint Paths and {\sf Minor Testing} on general graphs is to solve Disjoint Paths in single exponential time on surface-embedded graphs and in particular on planar graphs. Even when the inputs to Disjoint Paths are restricted to planar graphs, a case called the Planar Disjoint Paths problem, the best known algorithm has running time $2^{2^{O(k)}}n^2$. In this paper, we make the first step towards our quest for designing a single exponential time algorithm for Disjoint Paths by giving a $2^{O(k^2)}n^{O(1)}$-time algorithm for Planar Disjoint Paths.
翻译:在 Disjoint 路径问题中, 输入是一个没有方向的图形 $G$ 的 obson 和一套 $k 的 Overtices 和一套 $k 的 Overdex 配对, $_ i, t_ i i i = 1 ⁇ k 美元, 任务在于找到 $k 的双向的 vertex- dicomit 连接 $t$ 。 问题表现在 Robertson 和 Seymelmour 配对 。 现代术语中, 这意味着 Disjod 路径是固定的 podrable road (FT) 的 $G$ G$ G$ 。 这个算法中, 以 commodeal 运算算算算算方式中, Snational_ droadald 算算算算算中, 以 $ $x max max max 。