Scholarlink

<p>To improve helicopter flight safety in low-altitude urban environments under dynamic wind fields, this paper proposes a novel adaptive path-planning framework that integrates offline training with online planning. In the offline stage, deep neural network (DNN) surrogate models are trained using high-fidelity data to enable rapid wind-field reconstruction and flight-response prediction. In the online stage, a hybrid Kalman/CFD-DNN predictor is employed to estimate inflow wind speed and direction and reconstruct the urban wind fields. A dynamically updated safety map is then generated by evaluating helicopter control and attitude responses and by integrating typical threat zones around isolated buildings with globally distributed threat points. On this basis, an improved A* algorithm with dynamic safety mapping (I-DSM-A*) is developed for path-planning, with emphasis on route safety while maintaining computational efficiency. The results show that the proposed framework effectively mitigates threats induced by building wakes, reducing the number of search nodes and the computation time by nearly 50% under static wind conditions. In dynamic wind scenarios, the average time required for a single flow-field reconstruction and threat-identification cycle is 260&#xa0;ms, and the complete path-planning process is consistently completed within seconds. By balancing path safety, computational efficiency, and real-time performance, our method provides a practical and adaptive solution for ensuring the safe flight of helicopters operating in densely built urban environments.</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

Data-Driven Adaptive Safe Path-Planning for Helicopter in Urban Wind Fields

  • 期刊论文

  • 发表日期：

    2026年5月9日

查看全文
  • Yang Liu,
  • Yongjie Shi,
  • Yueyang Zhang,
  • Guohua Xu,
  • Weihong Kong

摘要

To improve helicopter flight safety in low-altitude urban environments under dynamic wind fields, this paper proposes a novel adaptive path-planning framework that integrates offline training with online planning. In the offline stage, deep neural network (DNN) surrogate models are trained using high-fidelity data to enable rapid wind-field reconstruction and flight-response prediction. In the online stage, a hybrid Kalman/CFD-DNN predictor is employed to estimate inflow wind speed and direction and reconstruct the urban wind fields. A dynamically updated safety map is then generated by evaluating helicopter control and attitude responses and by integrating typical threat zones around isolated buildings with globally distributed threat points. On this basis, an improved A* algorithm with dynamic safety mapping (I-DSM-A*) is developed for path-planning, with emphasis on route safety while maintaining computational efficiency. The results show that the proposed framework effectively mitigates threats induced by building wakes, reducing the number of search nodes and the computation time by nearly 50% under static wind conditions. In dynamic wind scenarios, the average time required for a single flow-field reconstruction and threat-identification cycle is 260 ms, and the complete path-planning process is consistently completed within seconds. By balancing path safety, computational efficiency, and real-time performance, our method provides a practical and adaptive solution for ensuring the safe flight of helicopters operating in densely built urban environments.