Analysing Adaption Processes of Hornets

In our research we study self-adapting multi-agent systems in the context of organisation-oriented design combined with the MAPE pattern (monitor, analyse, plan, and execute). We advocate a models@run.time approach for the planning phase to estimate the cost-benefit ratio of possible self-modifications. In this paper we develop key values to describe the dynamics of self-adapting systems. Our main objective is to compare adaption dynamics of similar systems, i.e., systems that differ only slightly, e.g., w.r.t. their organisational networks. We specify the MAPE-based adaption in the formalism of Hornets – a formalism that uses nets as tokens, i.e., they follow the nets-within-nets approach. We identify abstractions of the reachability graph that focus on the adaption aspects of the processes. We develop key measures for adaption processes on these abstracted graphs. The key measures are used e.g., to compare two variations of the same adapting system. The approach is illustrated by a case study: We analyse a Hornet-model of Axelrod’s well-known tournament where the playing agents adapt their strategies during the game.