On partially ordered graph grammars

Abstract

Graph grammars are systems for the generation of directed, node and edge labeled graphs. They rewrite single nodes only and establish connections between the inserted graph and the neighbors of the replaced node on the basis of node labels and edge labels. If there is only a single edge label, then graph grammars are closely related to NLC graph grammars.A partially ordered graph is a graph together with a spanning tree. These components are distinguished by their edge labels. A partially ordered graph grammar is the union of a graph grammar and a tree grammar. These components fit together such that their rewriting processes yield partially ordered graphs with the tree grammar generating spanning trees.Here we concentrate on the computational complexity of some restricted types of graph grammars and their languages with emphasis on intractability. It turns out that node and edge labeled tree grammars generate PSPACE-complete sets of connected graphs of finite degree, and that one-sided linear edge-unlabeled tree grammars generate NP-complete sets of graphs. However, the complexity is polynomial, if the graphs have finite degree and are generated by a one-sided linear partially ordered graph grammar. This situation closely parallels the case of NLC and regular BNLC grammars. NLC graph grammars can be seen as undirected, edge-unlabeled graph grammars, and, on the other hand, edge-unlabeled undirected one-sided linear partially ordered graph grammars and edge-unlabeled undirected one-sided linear partially ordered tree-graph grammars are special BNLC graph grammars.