Scholarlink

<p>With the rapid development of science and technology, unmanned aerial vehicles (UAVs) are widely used in modern battlefields, making effective judgment of their operational intent increasingly important. However, most existing research relies on rule-based and correlation analysis methods, which often assume independent and identically distributed data and thus perform poorly in adversarial, non-IID environments. Current machine learning approaches also face limitations in causal interpretability, cross-domain generalization, and robustness under noise. To address these issues, this paper proposes a Causal Inference-Based Human–Machine Hybrid Intelligence Model (CI-BNN-HM) for operational intent judgment. The model distinguishes causal and non-causal factors using a Structural Causal Model (SCM) and enhances feature selection with a mask module that combines dynamic selection, attention, and differentiable sampling. Furthermore, a human–machine hybrid framework based on Bayesian deep learning integrates complementary strengths of human and machine intelligence, improving accuracy under non-IID data. Comprehensive experiments show that the proposed method achieves 96.4% accuracy in UAV operational intent judgment. The mask module effectively reduces irrelevant features, yielding about 2% accuracy improvement compared to not selecting causal features. Introducing gating strategies to quantify human contributions further improves accuracy by about 2%. These results validate the model’s advantages in causal sufficiency, generalization, and robustness.</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

Human–Machine Hybrid Intelligence for Operational Intent Judgment: A Causal Inference Approach

  • 期刊论文

  • 发表日期：

    2025年12月12日

查看全文
  • Feng Chen,
  • Jin Hu,
  • Yinlong Yuan,
  • Wentao Chen,
  • Xiang Liu,
  • Yun Cheng,
  • Liang Hua

摘要

With the rapid development of science and technology, unmanned aerial vehicles (UAVs) are widely used in modern battlefields, making effective judgment of their operational intent increasingly important. However, most existing research relies on rule-based and correlation analysis methods, which often assume independent and identically distributed data and thus perform poorly in adversarial, non-IID environments. Current machine learning approaches also face limitations in causal interpretability, cross-domain generalization, and robustness under noise. To address these issues, this paper proposes a Causal Inference-Based Human–Machine Hybrid Intelligence Model (CI-BNN-HM) for operational intent judgment. The model distinguishes causal and non-causal factors using a Structural Causal Model (SCM) and enhances feature selection with a mask module that combines dynamic selection, attention, and differentiable sampling. Furthermore, a human–machine hybrid framework based on Bayesian deep learning integrates complementary strengths of human and machine intelligence, improving accuracy under non-IID data. Comprehensive experiments show that the proposed method achieves 96.4% accuracy in UAV operational intent judgment. The mask module effectively reduces irrelevant features, yielding about 2% accuracy improvement compared to not selecting causal features. Introducing gating strategies to quantify human contributions further improves accuracy by about 2%. These results validate the model’s advantages in causal sufficiency, generalization, and robustness.