Entropy-based goal-oriented emergence management in self-organizing systems through feedback control loop: A case study in NASA ANTS mission

Highlights: • Providing a new definition for emergence that is useful in guided artificial self-organizing systems. • Designing a new, MAPE-K based control loop for goal-oriented emergence management. • Providing an entropy-based algorithm for detection and management of emergence. • Improving availability in NASA ANTS case study using an entropy-based approach. The vast number of agents in artificial systems and maintenance issues have persuaded researchers to study self-organization. As a result of self-organization, some behaviors and structures emerge at a global level of the system. These emergents are created spontaneously and without a preexisting blueprint. Therefore, in some cases, they conflict with system goals (such as availability) in artificial self-organizing systems. So, many researchers' attention has been attracted to designing guided self-organizing systems and managing emergence. In this paper, a feedback control loop (inspired by IBM's control loop and relying on our proposed definition of emergence) is designed to manage emergence to achieve system goals. The emergence detection and management process is performed periodically by evaluating some agents' entropy in the system. Experiments on the NASA ANTS case study show the effectiveness of the proposed emergence management unit (called EMU). The results indicate that at all radio ranges considered, the entropy increment percentage when the EMU is enabled is much less than when the EMU is disabled. Besides, the availability of ruler agents using EMU is equal to or higher than the availability without using EMU.