Scholarlink

<p>Unmanned aerial vehicles (UAVs) are widely used in civilian and industrial applications owing to their versatility and autonomy. Safe automatic landing of fixed-wing UAVs is challenging, as the landing phase is sensitive to strong disturbances. To address this challenge, we propose an end-to-end anti-disturbance control framework. The framework integrates an Improved Soft Actor-Critic (SAC) algorithm with a Proportional-Integral-Derivative (PID) controller. The improved SAC agent generates control commands at the decision layer, which are then refined and executed by the PID controller to guarantee stable actuation. The improvement lies in incorporating a Curiosity Cross-Entropy Method (CCEM) into SAC, which improves sample efficiency and action quality by combining curiosity-based exploration with cross-entropy optimization. In addition, a multi-stage reward function is designed to ensure precise path following during the final landing phase. We further develop a high-fidelity simulation environment that models both external disturbances and internal failures. External disturbances include turbulence and Ornstein-Uhlenbeck (OU) noise, while internal disturbances simulate engine failures. Extensive experiments under diverse disturbance scenarios demonstrate that the proposed end-to-end control framework achieves robust and accurate path following, highlighting its potential for reliable automatic landing in emergency situations.</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

Improved SAC-PID path-following control of fixed-wing UAVs under engine failure

  • 期刊论文

  • 发表日期：

    2026年5月26日

查看全文
  • Qiaoli Zhou,
  • Bin Guo,
  • Tao Hong,
  • Zhiqiang Chang

摘要

Unmanned aerial vehicles (UAVs) are widely used in civilian and industrial applications owing to their versatility and autonomy. Safe automatic landing of fixed-wing UAVs is challenging, as the landing phase is sensitive to strong disturbances. To address this challenge, we propose an end-to-end anti-disturbance control framework. The framework integrates an Improved Soft Actor-Critic (SAC) algorithm with a Proportional-Integral-Derivative (PID) controller. The improved SAC agent generates control commands at the decision layer, which are then refined and executed by the PID controller to guarantee stable actuation. The improvement lies in incorporating a Curiosity Cross-Entropy Method (CCEM) into SAC, which improves sample efficiency and action quality by combining curiosity-based exploration with cross-entropy optimization. In addition, a multi-stage reward function is designed to ensure precise path following during the final landing phase. We further develop a high-fidelity simulation environment that models both external disturbances and internal failures. External disturbances include turbulence and Ornstein-Uhlenbeck (OU) noise, while internal disturbances simulate engine failures. Extensive experiments under diverse disturbance scenarios demonstrate that the proposed end-to-end control framework achieves robust and accurate path following, highlighting its potential for reliable automatic landing in emergency situations.