Label Constrained Shortest Path Algorithms: An Experimental Evaluation using Transportation Networks

We consider a generalization of the point-to-point (and single-source) shortest path problem to instances where the shortest path must satisfy a formal language constraint. Given an alphabet Σ, a (directed) network G whose edges are weighted and Σ-labeled, and a regular grammar L ⊆ Σ, the Regular Language Constrained Shortest Path problem consists of finding a shortest path p in G complying with the additional constraint that l(p) ∈ L. Here l(p) denotes the unique word given by concatenating the Σ-labels of the edges along the path p. In this chapter, we summarize our recent results and present new theoretical and experimental results for the Regular Language Constrained Shortest problem. We also present extensions of several speedup techniques developed earlier for the standard point-to-point shortest path problem. These speed-up techniques are integrated within the basic algorithmic framework to yield new algorithms for the problem. In order to evaluate the performance of the basic algorithm and its extensions, we have performed preliminary experimental analysis. Through our experiments, we study the scalability of the algorithm with respect to the network size as well as with respect to the constraining language complexity. Further, we study the effectiveness of speed-up techniques such as goal-directed and bidirectional search when applied to the Regular Language Constrained Shortest problem.