Scholarlink

<p>This article addresses the design problem of fractional order PID (FOPID) controllers in environments with disturbances and uncertainties for fixed-wing unmanned aerial vehicles (UAVs). While traditional FOPID controllers offer stable tracking, their performance degrades under dynamic external and internal interferences. To overcome this limitation, we propose a novel hybrid intelligent control strategy, termed PPO-based FOPID, which synergistically combines the stability of control theory with the adaptive learning capabilities of reinforcement learning. In this framework, an FOPID controller provides a baseline control policy, ensuring stability and tracking efficiency. Concurrently, a proximal policy optimization agent acts as a supplementary learning controller, continuously optimizing control commands via an actor-critic mechanism to actively counteract disturbances and uncertainties. The proposed strategy is validated on the attitude control system of a fixed-wing UAV. Simulation results validate the proposed controller’s superior performance. In experiments against conventional PID, FOPID, and an advanced fuzzy PID plus PID hybrid strategy, our method consistently achieves the lowest tracking error. It improves the root mean square error by 3.1<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(-\)</EquationSource> </InlineEquation>4.3% under external disturbances and up to 8.1% under parametric uncertainties, demonstrating significantly enhanced robustness and adaptability for high-precision UAV control in complex scenarios.</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 fractional-order PID controller design for fixed-wing UAVs through proximal-policy-optimization for disturbance rejection

  • 期刊论文

  • 发表日期：

    2025年10月9日

查看全文
  • Fei Xie,
  • Yi Chai,
  • Xiaolong Chen,
  • Ke Zhang,
  • Qie Liu

摘要

This article addresses the design problem of fractional order PID (FOPID) controllers in environments with disturbances and uncertainties for fixed-wing unmanned aerial vehicles (UAVs). While traditional FOPID controllers offer stable tracking, their performance degrades under dynamic external and internal interferences. To overcome this limitation, we propose a novel hybrid intelligent control strategy, termed PPO-based FOPID, which synergistically combines the stability of control theory with the adaptive learning capabilities of reinforcement learning. In this framework, an FOPID controller provides a baseline control policy, ensuring stability and tracking efficiency. Concurrently, a proximal policy optimization agent acts as a supplementary learning controller, continuously optimizing control commands via an actor-critic mechanism to actively counteract disturbances and uncertainties. The proposed strategy is validated on the attitude control system of a fixed-wing UAV. Simulation results validate the proposed controller’s superior performance. In experiments against conventional PID, FOPID, and an advanced fuzzy PID plus PID hybrid strategy, our method consistently achieves the lowest tracking error. It improves the root mean square error by 3.1 \(-\) 4.3% under external disturbances and up to 8.1% under parametric uncertainties, demonstrating significantly enhanced robustness and adaptability for high-precision UAV control in complex scenarios.