9 / 2024-06-11 22:57:38

Reinforcement learning-based resilience optimization of combat system-of-systems considering protective resource supplementation: From the perspective of continuous interference

Combat system-of-systems,Resilience optimization,Reinforcement learning,Protective resource supplement,Continuous interference

五、体系试验与评估 > 8.体系韧性评估与优化

As a complex military system composed of multiple interrelated and interdependent subsystems, combat system-of-systems (CSoS) plays an increasingly important role in modern warfare. However, continuous interference usually occurs to affect CSoS's normal services or functions after systems are deployed. Optimizing the resilience of CSoS holds significant value in minimizing service disruption, reducing adverse impact, and enhancing network survivability. From the perspective of continuous interference, we propose a united framework called reinforcement learning-based resilience optimization of CSoS considering protective resource supplementation (ROCSoS). Firstly, we present an index named the CSoS resilience index (CSoSRI) to measure the resilience of CSoS. This index integrates CSoS's structure, performance, performance baseline, as well as resistance and recovery capabilities in continuous interference. Subsequently, a resilience-oriented protection optimization method based on reinforcement learning (RL) is proposed by considering the dynamic nature of supplementation and the coexistence of interference, protection, and recovery. Finally, we conduct extensive simulation experiments on a CSoS case to demonstrate the effectiveness and superiority of the proposed ROCSoS. This study compares the ROCSoS with several baseline methods, revealing its superior performance. The results provide useful insights for guiding and designing the more resilient CSoS.