Core Concepts
Given a graph G, integers k and ℓ, and vertices s and t, the goal is to efficiently determine whether there exist k pairwise internally vertex-disjoint s-t paths of length at most ℓ.
Abstract
The Short Path Packing (SPP) problem is a fundamental graph problem with applications in network survivability and graph sparsity measures. It has been shown to be NP-hard for k ≥ 2 and ℓ ≥ 5.
The key insights and steps of the approach are:
Greedy Approach:
A greedy algorithm can find disjoint short paths, but it fails to determine whether a solution exists when it cannot find k such paths.
The failure of the greedy algorithm indicates that the solution paths must intersect the previously computed paths at some internal vertices.
Checkpoints:
To exploit this insight, the problem is reformulated as Short Path Packing with Checkpoints (SPPC), where the terminal pairs are replaced by terminal lists with intermediate checkpoints.
Two types of failures can occur in the greedy algorithm for SPPC: Failure Condition 1 (FC1) when no path exists between two checkpoints, and Failure Condition 2 (FC2) when the total length of the computed paths exceeds the length bound ℓ.
Lemmas 3.6 and 3.7 show that in both cases, the solution paths must use some vertices that were already used by the greedy algorithm.
Search-Tree Algorithm:
The algorithm uses a search-tree approach, where the greedy algorithm is first applied, and in case of failure, the search tree is explored by branching on the vertices that must be used by the solution paths.
The search-tree algorithm has a worst-case runtime complexity of (k · ℓ^2)^(k·ℓ) · n^O(1), which is larger than the state-of-the-art FPT algorithm, but allows for a broader range of potential optimizations.