Planning Self-Adaption with Graph Transformations

Paper in proceeding, 2012

Self-adaptive systems autonomously adjust their behavior in order to achieve their goals despite changes in the environment and the system itself. Self-adaption is typically implemented in software and often expressed in terms of architectural reconfiguration. The graph transformation formalism is a natural way to model the architectural reconfiguration in self-adaptive systems. In this paper, we present (1) how we employ graph transformations for the specification of architectural reconfiguration and (2) how we transform graph transformations into actions of the Planning Domain Definition Language (PDDL) in order to use off-the-shelf tools for the computation of self-adaptation plans. We illustrate our approach by a self-healing process and show the results of a simulation case study.