Scholarlink

<p>Motion planning is a critical module in autonomous driving, and as an essential approach within it, spatio-temporal planners effectively handle dynamic scenarios but still face challenges in achieving high-quality solutions. Meta-heuristic optimization, a class of artificial intelligence algorithms, is widely used for complex optimization problems. This paper proposes a Multi-strategy Enhanced Snake Optimizer (MESO) for motion planning in autonomous vehicles. MESO integrates three key enhancements: (1) a hybrid exploration strategy combining Roulette Wheel Selection (RWS) and Lévy flight to improve search diversity, (2) an adaptive parameter update mechanism based on information entropy for enhanced convergence stability, and (3) an improved egg hatching strategy using Evolutionary Population Dynamics (IEPD) to strengthen local exploitation. Extensive evaluations on 23 benchmark functions and the CEC2022 suite demonstrate MESO’s effectiveness, ranking first in 6 out of 12 functions in the CEC2022 suite and in 21 out of 23 benchmark functions, ultimately achieving the highest overall ranking across both tests. Motion planning simulation test further demonstrate MESO’s superiority over the original Snake Optimizer, consistently achieving higher-quality solutions. Specifically, MESO improved the average objective value, standard deviation, and average convergence time by 23.2%, 87.4%, and 56.1%, respectively, in static scenarios, and by 18.6%, 83.6%, and 45.1%, respectively, in dynamic scenarios. Finally, hardware-in-the-loop (HIL) testing on a real vehicle model verifies MESO’s feasibility for autonomous driving applications. The source code of MESO is currently available for public from: <a href="https://github.com/Qilinn-Robotics/MESO-2025">https://github.com/Qilinn-Robotics/MESO-2025</a>.</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

MESO: a multi-strategy enhanced snake optimizer applied to autonomous vehicle motion planning

  • 期刊论文

  • 发表日期：

    2025年8月30日

查看全文
  • Qilin Li,
  • Chunyan Zhang,
  • Qihua Ma,
  • Xin Weng

摘要

Motion planning is a critical module in autonomous driving, and as an essential approach within it, spatio-temporal planners effectively handle dynamic scenarios but still face challenges in achieving high-quality solutions. Meta-heuristic optimization, a class of artificial intelligence algorithms, is widely used for complex optimization problems. This paper proposes a Multi-strategy Enhanced Snake Optimizer (MESO) for motion planning in autonomous vehicles. MESO integrates three key enhancements: (1) a hybrid exploration strategy combining Roulette Wheel Selection (RWS) and Lévy flight to improve search diversity, (2) an adaptive parameter update mechanism based on information entropy for enhanced convergence stability, and (3) an improved egg hatching strategy using Evolutionary Population Dynamics (IEPD) to strengthen local exploitation. Extensive evaluations on 23 benchmark functions and the CEC2022 suite demonstrate MESO’s effectiveness, ranking first in 6 out of 12 functions in the CEC2022 suite and in 21 out of 23 benchmark functions, ultimately achieving the highest overall ranking across both tests. Motion planning simulation test further demonstrate MESO’s superiority over the original Snake Optimizer, consistently achieving higher-quality solutions. Specifically, MESO improved the average objective value, standard deviation, and average convergence time by 23.2%, 87.4%, and 56.1%, respectively, in static scenarios, and by 18.6%, 83.6%, and 45.1%, respectively, in dynamic scenarios. Finally, hardware-in-the-loop (HIL) testing on a real vehicle model verifies MESO’s feasibility for autonomous driving applications. The source code of MESO is currently available for public from: https://github.com/Qilinn-Robotics/MESO-2025.