Graph Matching for Context Recognition

In the software implementation of a general Ambient Intelligence (AmI) system, there are two major issues, on which depend the flexibility and the performance of the project. One is the implementation paradigm - how the various entities are organized and how they interact; the other is the management of context information, and how context-awareness is integrated as a first-class element in the implementation. This paper is framed in a research effort to develop an agent-based platform for AmI applications. While in previous research we have already argued in favor of using an agent-oriented paradigm for the implementation, and we have already introduced the concept of context graphs and context patterns, it is in this paper that we argue that matching context patters against context graphs is a valid method for detecting the user´s situation and acting upon the user´s context. In support of this, we analyze several algorithms for graph matching, adapted to our problem, and compare their performance on specific examples of context matching.