Scholarlink

<p>Unmanned weapon system-of-systems (UWSoS) require advanced resilience mechanisms to sustain mission-critical performance under adversarial conditions. This paper presents the meta-path-guided causal inference (MPGCI) framework, which seamlessly integrates meta-path-based feature extraction, causal disentanglement, and reinforcement learning to significantly enhance UWSoS resilience. The meta-path pool enables efficient neighbor sampling, effectively capturing complex higher-order network dependencies, while a directed acyclic graph aligns features with underlying causal structures, mitigating the effects of semantic drift. Furthermore, an actor-critic network, constrained by meta-paths, optimizes task allocation and recovery strategies in dynamic, evolving environments. By embedding meta-paths across the framework, MPGCI ensures precise feature alignment and adaptive resilience. Experimental results demonstrate MPGCI’s superior performance compared to state-of-the-art methods, highlighting its robustness in maintaining operational stability and facilitating rapid recovery in complex scenarios.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Meta-path-guided causal inference for hierarchical feature alignment and policy optimization in enhancing resilience of UWSoS

  • 期刊论文

  • 发表日期：

    2025年1月3日

查看全文
  • Kexin Wang,
  • Dingrui Xue,
  • Yingdong Gou,
  • Wanlong Qi,
  • Bo Li,
  • Jiancheng Liu,
  • Yinglong Feng,
  • Yuqing Lin

摘要

Unmanned weapon system-of-systems (UWSoS) require advanced resilience mechanisms to sustain mission-critical performance under adversarial conditions. This paper presents the meta-path-guided causal inference (MPGCI) framework, which seamlessly integrates meta-path-based feature extraction, causal disentanglement, and reinforcement learning to significantly enhance UWSoS resilience. The meta-path pool enables efficient neighbor sampling, effectively capturing complex higher-order network dependencies, while a directed acyclic graph aligns features with underlying causal structures, mitigating the effects of semantic drift. Furthermore, an actor-critic network, constrained by meta-paths, optimizes task allocation and recovery strategies in dynamic, evolving environments. By embedding meta-paths across the framework, MPGCI ensures precise feature alignment and adaptive resilience. Experimental results demonstrate MPGCI’s superior performance compared to state-of-the-art methods, highlighting its robustness in maintaining operational stability and facilitating rapid recovery in complex scenarios.