Matching Scenarios Patterns by Using Linear Programming

Abstract

This paper continues the work presented previously at ICNC-FSKD-05 for representing and matching scenario patterns. A unified scheme is presented to replace the two previous equivalent schemas for formalizing scenario patterns. In the unified scheme, a scenario is denoted in terms of a collection of states with the corresponding temporal constraints, where a state is defined as a set of Boolean-valued time-dependent fluents. The concept of a scenario graph is formally introduced as a directed, partially weighted and labeled simple graph. Based on such a graphical representation, an extended linear programming graph matching algorithm is proposed for recognizing scenario patterns.