计算机工程与应用 ›› 2023, Vol. 59 ›› Issue (14): 76-85.DOI: 10.3778/j.issn.1002-8331.2209-0332

• 理论与研发 • 上一篇    下一篇

融合牵引变异和透镜成像的花授粉算法及应用

李大海,伍兆前,王振东   

  1. 江西理工大学 信息工程学院,江西 赣州 341099
  • 出版日期:2023-07-15 发布日期:2023-07-15

Flower Pollination Algorithm Combining Lens Imaging and Traction Mutation and Its Application

LI Dahai, WU Zhaoqian, WANG Zhendong   

  1. School of Information Engineering, Jiangxi University of Science and Technology, Ganzhou, Jiangxi 341099, China
  • Online:2023-07-15 Published:2023-07-15

摘要: 针对花授粉算法收敛速度较慢、容易陷入局部最优解等问题,提出结合牵引变异和改进型透镜成像的增强花授粉算法LMFPA(lens learning and traction mutation based enhanced flower pollination algorithm)。LMFPA算法利用改进型透镜成像机制优化算法的种群分布。通过增加观测因子的牵引变异机制增加算法收敛速度和跳出局部最优的能力。以CEC2013中的12个测试函数作为性能评测函数集,将LMFPA与FPA(follower pollination algorithm)、TMFPA(T-distribution mutation-based flower pollination algorithm)、t-GSSA(improved sparrow search algorithm based on adaptive t-distribution and golden sine and its application)和PCSPSO(particle compaction and scheduling based particle swarm optimization)4个改进型FPA算法进行评测。实验结果表明LMFPA算法无论是收敛速度还是收敛精度上都占优。将LMFPA应用于在无人机三维路径规划问题,实验结果表明LMFPA也能取得更优的三维路径规划结果。

关键词: 花授粉算法, 透镜成像, 牵引变异, 路径规划

Abstract: Due to defects of relatively slow convergence and being easily trapped in local optimal of flower pollination algorithm(FPA), an enhanced lens learning and traction mutation based flower pollination algorithm(LMFPA) is proposed in this paper. First, LMFPA applies a modified lens learning mechanism to improve the distribution of population. Second, LMFPA uses the observation factor based traction mutation strategy to further accelerate the convergence and increase the probability to jumping out of local optimal. 12 benchmark functions from CEC2013 are selected as testbed to evaluate performance of LMFPA with original FPA algorithm and another 3 improved FPA algorithms, including TMFPA (T-distribution mutation-based flower pollination algorithm), t-GSSA(improved sparrow search algorithm based on adaptive t-distribution and golden sine and its application) and PCSPSO(particle compaction and scheduling based particle swarm optimization). Experimental result shows that LMFPA can achieve superior performance both in convergence speed and accuracy. At last, LMFPA is also used to solve 3D path planning for UAVs. Experimental result illustrates that LMFPA can also find better 3D paths for UAVs.

Key words: flower pollination algorithm, lens imaging, traction variation, path planning