AI-based SoS performance classification for resilience reaction

Abstract

This works provides a comprehensive performance evaluation approach for system-of-systems (SoS). Various attributes are extracted from different level and various aspects, such as graph-based communication quality, dynamic response of physical component systems (PCSs) and mission-oriented features of management component systems (MCSs). AI-based classifier is employed thereafter based on the performance feature construction, to classify SoS performance into 5 status, Normal, Robust, Fault-tolerance, Resilience and Breakdown, which shows the necessity of a resilience SoS reaction. The effectiveness of the performance feature construction is verified by the high accuracy of training set in the simulation.