Scholarlink

<p>Unmanned aerial vehicles (UAVs) require robust path-following in high winds, yet accurate wind measurement is often impractical. This paper presents ANN-PLOS, a sensor-free adaptive guidance law that overcomes this limitation by employing compact artificial neural networks (ANNs) to dynamically optimize the gains of a pursuit line-of-sight (PLOS) guidance law in real time. The ANNs are trained offline using the tabu continuous ant colony system algorithm, minimizing a combined cost of cross-track error and control effort. Crucially, the framework guarantees stable and bounded tracking under turbulent wind disturbances without requiring wind sensors or complex estimators. In simulation, ANN-PLOS outperforms conventional fixed-gain PLOS by 22% in tracking accuracy, with faster convergence and lower energy consumption. It also surpasses state-of-the-art adaptive methods in control smoothness and computational efficiency, which is critical for resource-constrained UAVs. The results demonstrate that ANN-PLOS provides a rigorous, lightweight, and practically viable solution for reliable UAV navigation in dynamically windy 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

Robust real-time UAV path-following method in high winds via compact neural networks

  • 期刊论文

  • 发表日期：

    2026年4月22日

查看全文
  • Javad Asghari Jigheh,
  • Morteza Tayefi

摘要

Unmanned aerial vehicles (UAVs) require robust path-following in high winds, yet accurate wind measurement is often impractical. This paper presents ANN-PLOS, a sensor-free adaptive guidance law that overcomes this limitation by employing compact artificial neural networks (ANNs) to dynamically optimize the gains of a pursuit line-of-sight (PLOS) guidance law in real time. The ANNs are trained offline using the tabu continuous ant colony system algorithm, minimizing a combined cost of cross-track error and control effort. Crucially, the framework guarantees stable and bounded tracking under turbulent wind disturbances without requiring wind sensors or complex estimators. In simulation, ANN-PLOS outperforms conventional fixed-gain PLOS by 22% in tracking accuracy, with faster convergence and lower energy consumption. It also surpasses state-of-the-art adaptive methods in control smoothness and computational efficiency, which is critical for resource-constrained UAVs. The results demonstrate that ANN-PLOS provides a rigorous, lightweight, and practically viable solution for reliable UAV navigation in dynamically windy environments.