It is known that seemingly small details such as tie-breaking among nodes with the same f-cost can significantly affect the performance of a best-first search algorithm on many domains (Asai and Fukunaga 2017). In this paper, we show that low-level algorithmic details of domain-independent planning heuristics can have a surprisingly large impact on search performance. As a case study, we consider the well-known FF heuristic (hff) (Hoffmann and Nebel 2001).

We propose an exact algorithm for solving the longest path problem between two given vertices in undirected weighted graphs. By using graph partitioning and dynamic programming, we obtain an algorithm that is significantly faster than other state-of-the-art methods. This enables us to solve instances that were previously unsolved and solve hard instances significantly faster. We also present a parallel version of the algorithm.