Scholarlink

<p>In non-towered terminal airspace, the lack of real-time air traffic control makes accurate trajectory prediction a critical task for aviation safety. To address the challenges of trajectory prediction caused by frequent altitude changes in this airspace, we propose ACTrajNet, an altitude-aware trajectory prediction model. The model independently extracts altitude features using temporal convolutional networks(TCN), it then incorporates a channel attention fusion mechanism to dynamically fuse altitude features into the trajectory representation across different channels. This significantly enhances the model’s ability to capture complex vertical flight patterns. ACTrajNet outperforms state-of-the-art models on the TrajAir dataset, achieving up to 2.47% and 2.42% improvement in ADE(Average Displacement Error) and FDE(Final Displacement Error), respectively, compared to the baseline model TrajAirNet. These results demonstrate the potential of ACTrajNet to improve trajectory prediction under complex flight conditions.</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

Altitude aware trajectory prediction methods for non towered terminal airspace

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月23日

查看全文
  • Haipeng Zhu,
  • Qiang Tong,
  • Jinqing Hu,
  • Xiulei Liu,
  • Shoulu Hou

摘要

In non-towered terminal airspace, the lack of real-time air traffic control makes accurate trajectory prediction a critical task for aviation safety. To address the challenges of trajectory prediction caused by frequent altitude changes in this airspace, we propose ACTrajNet, an altitude-aware trajectory prediction model. The model independently extracts altitude features using temporal convolutional networks(TCN), it then incorporates a channel attention fusion mechanism to dynamically fuse altitude features into the trajectory representation across different channels. This significantly enhances the model’s ability to capture complex vertical flight patterns. ACTrajNet outperforms state-of-the-art models on the TrajAir dataset, achieving up to 2.47% and 2.42% improvement in ADE(Average Displacement Error) and FDE(Final Displacement Error), respectively, compared to the baseline model TrajAirNet. These results demonstrate the potential of ACTrajNet to improve trajectory prediction under complex flight conditions.