Enhancing the resilience of combat system-of-systems under continuous attacks: Novel index and reinforcement learning-based protection optimization

Abstract

Combat system-of-systems (CSoS) comprises diverse combat entities with distinct functions designed to meet future task requirements, and it serves as a central component of military reform. Enhancing the resilience of CSoS is of significant military value for optimizing combat architecture, ensuring military operational capabilities and improving military survivability. This paper proposes an integrated framework called CSoS capability resilience (CSoSCR) framework to enhance the resilience of the CSoS under continuous attacks. Specifically, a continuous attacks resilience index (CARI) is proposed to evaluate the resilience of the CSoS under continuous attacks considering the task baseline, adaptability, recovery probability, and continuous attacks characteristics. Based on this setup, focusing on the degenerate phase, a resilience-oriented protection optimization method based on reinforcement learning (PORL) is presented to identify nodes requiring protecting and then enhance the resilience of CSoS. Finally, the reliability, effectiveness, and superiority of the CSoSCR framework are tested through a comprehensive case study. The results offer valuable insights for operational guidance and decision support in designing more resilient CSoS.