Scholarlink

<p>Reinforcement learning (RL) methods face substantial challenges in unmanned aerial vehicle (UAV) charging path planning under the uncertain motion of mobile charging vehicles (MCVs), including low exploration efficiency, irregular motion patterns and large prediction errors. To address these issues, this paper proposes an adaptive SARSA algorithm for dynamic environments (ASDE). ASDE incorporates an MCV motion model with a restructured reward function. In addition, it employs a time-varying <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\varepsilon \)</EquationSource> <EquationSource Format="MATHML"><math> <mi>ε</mi> </math></EquationSource> </InlineEquation>-greedy strategy that adaptively balances exploration and exploitation according to environmental feedback. A hybrid temporal difference (TD) (<InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(\lambda \)</EquationSource> <EquationSource Format="MATHML"><math> <mi>λ</mi> </math></EquationSource> </InlineEquation>) multi-step backtracking mechanism further enables dynamic adjustment of the backtracking horizon based on the MCV’s instantaneous motion, thereby enhancing predictive accuracy and responsiveness. Simulation results indicate that ASDE achieves convergence approximately 45% faster than conventional SARSA. In environments ranging from 20 <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(\times \)</EquationSource> <EquationSource Format="MATHML"><math> <mo>×</mo> </math></EquationSource> </InlineEquation> 20 to 50 <InlineEquation ID="IEq4"> <EquationSource Format="TEX">\(\times \)</EquationSource> <EquationSource Format="MATHML"><math> <mo>×</mo> </math></EquationSource> </InlineEquation> 50, ASDE improves task success rates by 4.4–10.4%, shortens path lengths by 21.2–30.7%, and reduces inflection points by 33.0–46.0% compared with benchmark algorithms.</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

ASDE Algorithm-Based UAV Dynamic Charging Path Planning Method

  • 期刊论文

  • 发表日期：

    2026年7月14日

查看全文
  • Dan Shan,
  • Meng Zhang,
  • Jianwei He,
  • Tianyu Zhang,
  • Yanfeng Li

摘要

Reinforcement learning (RL) methods face substantial challenges in unmanned aerial vehicle (UAV) charging path planning under the uncertain motion of mobile charging vehicles (MCVs), including low exploration efficiency, irregular motion patterns and large prediction errors. To address these issues, this paper proposes an adaptive SARSA algorithm for dynamic environments (ASDE). ASDE incorporates an MCV motion model with a restructured reward function. In addition, it employs a time-varying \(\varepsilon \) ε -greedy strategy that adaptively balances exploration and exploitation according to environmental feedback. A hybrid temporal difference (TD) ( \(\lambda \) λ ) multi-step backtracking mechanism further enables dynamic adjustment of the backtracking horizon based on the MCV’s instantaneous motion, thereby enhancing predictive accuracy and responsiveness. Simulation results indicate that ASDE achieves convergence approximately 45% faster than conventional SARSA. In environments ranging from 20 \(\times \) × 20 to 50 \(\times \) × 50, ASDE improves task success rates by 4.4–10.4%, shortens path lengths by 21.2–30.7%, and reduces inflection points by 33.0–46.0% compared with benchmark algorithms.