Model-driven Engineering of Decentralized Control in Cyber-Physical Systems

Self-adaptation is nowadays recognized as an effective approach to manage the complexity and dynamics inherent to cyber-physical systems, which are composed of deeply intertwined physical and software components interacting with each other. A self-adaptive system typically consists of a managed subsystem and a managing subsystem that implements the adaptation logic by means of the well established MAPE-K control loop. Since in large distributed settings centralized control is hardly adequate to manage the whole system, self-adaptation should be achieved through collective decentralized control, that is multiple cyber-physical entities must adapt in order to address critical runtime conditions. Developing such systems is challenging, as several dimensions concerning both the cyber-physical system and the decentralized control loop should be considered. To this end, we promote MAPE-K components as first-class modeling abstractions and provide a framework supporting the design, development, and validation of decentralized self-adaptive cyber-physical systems.