Modelling and analysis of self-adaptive systems using petri nets : a robot kitchen case study

Abstract

Current cyber-physical systems (CPS) consist of connected sub-systems. Recently, flexibility and adaptivity gain more and more attention in application areas, e.g. in the smart factory context. Multi-Agent Systems (MAS) are used to model CPS sub-systems and their interaction. Recent research focuses on the self-organization of these CPS-MAS. In MAS the concept of an organization (and consequently that of self-organization) is a research topic on its own. Within our research, we address the modeling and the analysis of these self-organizing MAS-Organizations. As part of this paper, we demonstrate the usefulness of formal modeling techniques (here: high-level Petri Nets) to obtain qualitative and quantitative insights. Here, a robot kitchen is used as a metaphor for a flexible production scenario: We have recipes as production jobs, cooking robots with tools as interacting CPS, a conveyor belt as a transportation system for the dishes under preparation and so on. The scenario provides several sources of adaptivity: the robot cooks may decide to change the priority of jobs due to changing external signals (prices); the kitchen may run out of stock; a specialized robot may become a bottleneck and other robots change their tools to support it. Our model is implemented in RENEW, an interactive Petri net simulator that supports nets-within-nets, which is a formalism defined to support self-modification in a direct manner. The graphical model deepens the understanding of the application domain (the real world) among the project partners; the formal model allows for qualitative analysis, for example the absence of deadlocks, as well as quantitative aspects like finishing time and throughput.