Scholarlink

<p>A dual-layer path planning framework (DLPPF) is proposed to enhance the safety and robustness of autonomous underwater vehicles (AUVs) operating under navigation uncertainty in GPS-denied environments. In the upper layer, an error-aware Q-learning algorithm is employed for global route generation, where a three-dimensional navigation-error is introduced as a time-varying safety radius. This design enables the planner to produce globally feasible paths under uncertain navigation conditions. The upper planner provides a sequence of reference waypoints and corresponding safety margins to guide the local planner. In the lower layer, an Adaptive Dynamic Window Approach (ADWA) serves as the local planner, which dynamically adjusts the velocity sampling window according to the proximity of obstacles and the geometric relation to the target. By integrating adaptive scaling into the velocity evaluation process, the ADWA achieves improved path smoothness and stability while maintaining efficient real-time performance. Simulation results in a complex three-dimensional underwater environment (5&#xa0;km&#xa0;<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\times\)</EquationSource> </InlineEquation>&#xa0;5&#xa0;km&#xa0;<InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(\times\)</EquationSource> </InlineEquation>&#xa0;2.5&#xa0;km) demonstrate that the proposed framework generates reliable global paths under different navigation-error conditions and enables the ADWA to achieve more effective real-time obstacle avoidance. Compared with conventional DWA and MPC planners, the ADWA yields 6–17% shorter paths across various obstacle configurations, provides noticeable energy savings in single-obstacle scenarios, and produces smoother trajectories with larger effective safety margins.</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

Dual-layer path planning for autonomous underwater vehicles under navigation uncertainty

  • 期刊论文

  • 发表日期：

    2026年3月30日

查看全文
  • Haomiao Yu,
  • Yuhe Dou,
  • Yulong Tuo

摘要

A dual-layer path planning framework (DLPPF) is proposed to enhance the safety and robustness of autonomous underwater vehicles (AUVs) operating under navigation uncertainty in GPS-denied environments. In the upper layer, an error-aware Q-learning algorithm is employed for global route generation, where a three-dimensional navigation-error is introduced as a time-varying safety radius. This design enables the planner to produce globally feasible paths under uncertain navigation conditions. The upper planner provides a sequence of reference waypoints and corresponding safety margins to guide the local planner. In the lower layer, an Adaptive Dynamic Window Approach (ADWA) serves as the local planner, which dynamically adjusts the velocity sampling window according to the proximity of obstacles and the geometric relation to the target. By integrating adaptive scaling into the velocity evaluation process, the ADWA achieves improved path smoothness and stability while maintaining efficient real-time performance. Simulation results in a complex three-dimensional underwater environment (5 km  \(\times\)  5 km  \(\times\)  2.5 km) demonstrate that the proposed framework generates reliable global paths under different navigation-error conditions and enables the ADWA to achieve more effective real-time obstacle avoidance. Compared with conventional DWA and MPC planners, the ADWA yields 6–17% shorter paths across various obstacle configurations, provides noticeable energy savings in single-obstacle scenarios, and produces smoother trajectories with larger effective safety margins.