Scholarlink

<p>To address security threats such as signal interference and atypical data injection in modern aerial systems, this paper proposes an aircraft trajectory anomaly identification method based on Multimodal Large Language Models (MLLMs). Traditional statistical “outlier detection” often fails to distinguish between complex flight maneuvers and stealthy anomalies. To overcome this, we first construct a high-fidelity 3D trajectory simulation dataset based on aerodynamic and kinematic mechanisms across three flight mission profiles. A physics-based framework is then developed to simulate five typical anomaly patterns, including oscillatory noise, positional drift, and velocity perturbation. By leveraging the visual perception and logical reasoning of MLLMs, combined with Chain-of-Thought (CoT) and prompt engineering, trajectory data is transformed into “vision-text” multimodal inputs. Experimental results demonstrate that the proposed method achieves an overall accuracy of 89.73%, a detection rate of 83.29%, and a false alarm rate of 7.52%. Compared with traditional Isolation Forest and Transformer baselines, the proposed approach improves accuracy by nearly 25%, effectively reduces false positives in complex maneuver scenarios, and achieves effective identification of deceptive anomalies with physical interpretability. Nevertheless, the results also indicate that slow progressive drift anomalies, such as altitude deviation and velocity perturbation, remain more challenging and require further sensitivity optimization.</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

Multimodal large language model-based trajectory anomaly detection in aerial systems

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月24日

查看全文
  • Yunduo Wu,
  • Lanyu Li,
  • Yuhao Yang,
  • Beile Niu,
  • Shiming Huang,
  • Xiaogang Yu

摘要

To address security threats such as signal interference and atypical data injection in modern aerial systems, this paper proposes an aircraft trajectory anomaly identification method based on Multimodal Large Language Models (MLLMs). Traditional statistical “outlier detection” often fails to distinguish between complex flight maneuvers and stealthy anomalies. To overcome this, we first construct a high-fidelity 3D trajectory simulation dataset based on aerodynamic and kinematic mechanisms across three flight mission profiles. A physics-based framework is then developed to simulate five typical anomaly patterns, including oscillatory noise, positional drift, and velocity perturbation. By leveraging the visual perception and logical reasoning of MLLMs, combined with Chain-of-Thought (CoT) and prompt engineering, trajectory data is transformed into “vision-text” multimodal inputs. Experimental results demonstrate that the proposed method achieves an overall accuracy of 89.73%, a detection rate of 83.29%, and a false alarm rate of 7.52%. Compared with traditional Isolation Forest and Transformer baselines, the proposed approach improves accuracy by nearly 25%, effectively reduces false positives in complex maneuver scenarios, and achieves effective identification of deceptive anomalies with physical interpretability. Nevertheless, the results also indicate that slow progressive drift anomalies, such as altitude deviation and velocity perturbation, remain more challenging and require further sensitivity optimization.